Commit | Line | Data |
---|---|---|
8e854e9c GR |
1 | /* |
2 | * Copyright (C) 2015, SUSE | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU General Public License as published by | |
6 | * the Free Software Foundation; either version 2, or (at your option) | |
7 | * any later version. | |
8 | * | |
9 | */ | |
10 | ||
11 | ||
12 | #include <linux/module.h> | |
47741b7c GR |
13 | #include <linux/dlm.h> |
14 | #include <linux/sched.h> | |
1aee41f6 | 15 | #include <linux/raid/md_p.h> |
47741b7c | 16 | #include "md.h" |
e94987db | 17 | #include "bitmap.h" |
edb39c9d | 18 | #include "md-cluster.h" |
47741b7c GR |
19 | |
20 | #define LVB_SIZE 64 | |
1aee41f6 | 21 | #define NEW_DEV_TIMEOUT 5000 |
47741b7c GR |
22 | |
23 | struct dlm_lock_resource { | |
24 | dlm_lockspace_t *ls; | |
25 | struct dlm_lksb lksb; | |
26 | char *name; /* lock name. */ | |
27 | uint32_t flags; /* flags to pass to dlm_lock() */ | |
47741b7c | 28 | struct completion completion; /* completion for synchronized locking */ |
c4ce867f GR |
29 | void (*bast)(void *arg, int mode); /* blocking AST function pointer*/ |
30 | struct mddev *mddev; /* pointing back to mddev. */ | |
31 | }; | |
32 | ||
96ae923a GR |
33 | struct suspend_info { |
34 | int slot; | |
35 | sector_t lo; | |
36 | sector_t hi; | |
37 | struct list_head list; | |
38 | }; | |
39 | ||
40 | struct resync_info { | |
41 | __le64 lo; | |
42 | __le64 hi; | |
43 | }; | |
44 | ||
fa8259da GR |
45 | /* md_cluster_info flags */ |
46 | #define MD_CLUSTER_WAITING_FOR_NEWDISK 1 | |
90382ed9 | 47 | #define MD_CLUSTER_SUSPEND_READ_BALANCING 2 |
eece075c | 48 | #define MD_CLUSTER_BEGIN_JOIN_CLUSTER 3 |
fa8259da GR |
49 | |
50 | ||
c4ce867f GR |
51 | struct md_cluster_info { |
52 | /* dlm lock space and resources for clustered raid. */ | |
53 | dlm_lockspace_t *lockspace; | |
cf921cc1 GR |
54 | int slot_number; |
55 | struct completion completion; | |
c4ce867f | 56 | struct mutex sb_mutex; |
54519c5f | 57 | struct dlm_lock_resource *bitmap_lockres; |
96ae923a GR |
58 | struct list_head suspend_list; |
59 | spinlock_t suspend_lock; | |
e94987db GR |
60 | struct md_thread *recovery_thread; |
61 | unsigned long recovery_map; | |
4664680c GR |
62 | /* communication loc resources */ |
63 | struct dlm_lock_resource *ack_lockres; | |
64 | struct dlm_lock_resource *message_lockres; | |
65 | struct dlm_lock_resource *token_lockres; | |
1aee41f6 | 66 | struct dlm_lock_resource *no_new_dev_lockres; |
4664680c | 67 | struct md_thread *recv_thread; |
1aee41f6 | 68 | struct completion newdisk_completion; |
fa8259da | 69 | unsigned long state; |
4664680c GR |
70 | }; |
71 | ||
72 | enum msg_type { | |
73 | METADATA_UPDATED = 0, | |
74 | RESYNCING, | |
1aee41f6 | 75 | NEWDISK, |
88bcfef7 | 76 | REMOVE, |
97f6cd39 | 77 | RE_ADD, |
dc737d7c | 78 | BITMAP_NEEDS_SYNC, |
4664680c GR |
79 | }; |
80 | ||
81 | struct cluster_msg { | |
82 | int type; | |
83 | int slot; | |
1aee41f6 | 84 | /* TODO: Unionize this for smaller footprint */ |
4664680c GR |
85 | sector_t low; |
86 | sector_t high; | |
1aee41f6 GR |
87 | char uuid[16]; |
88 | int raid_slot; | |
47741b7c GR |
89 | }; |
90 | ||
91 | static void sync_ast(void *arg) | |
92 | { | |
93 | struct dlm_lock_resource *res; | |
94 | ||
95 | res = (struct dlm_lock_resource *) arg; | |
96 | complete(&res->completion); | |
97 | } | |
98 | ||
99 | static int dlm_lock_sync(struct dlm_lock_resource *res, int mode) | |
100 | { | |
101 | int ret = 0; | |
102 | ||
47741b7c GR |
103 | ret = dlm_lock(res->ls, mode, &res->lksb, |
104 | res->flags, res->name, strlen(res->name), | |
105 | 0, sync_ast, res, res->bast); | |
106 | if (ret) | |
107 | return ret; | |
108 | wait_for_completion(&res->completion); | |
109 | return res->lksb.sb_status; | |
110 | } | |
111 | ||
112 | static int dlm_unlock_sync(struct dlm_lock_resource *res) | |
113 | { | |
114 | return dlm_lock_sync(res, DLM_LOCK_NL); | |
115 | } | |
116 | ||
c4ce867f | 117 | static struct dlm_lock_resource *lockres_init(struct mddev *mddev, |
47741b7c GR |
118 | char *name, void (*bastfn)(void *arg, int mode), int with_lvb) |
119 | { | |
120 | struct dlm_lock_resource *res = NULL; | |
121 | int ret, namelen; | |
c4ce867f | 122 | struct md_cluster_info *cinfo = mddev->cluster_info; |
47741b7c GR |
123 | |
124 | res = kzalloc(sizeof(struct dlm_lock_resource), GFP_KERNEL); | |
125 | if (!res) | |
126 | return NULL; | |
b83d51c0 | 127 | init_completion(&res->completion); |
c4ce867f GR |
128 | res->ls = cinfo->lockspace; |
129 | res->mddev = mddev; | |
47741b7c GR |
130 | namelen = strlen(name); |
131 | res->name = kzalloc(namelen + 1, GFP_KERNEL); | |
132 | if (!res->name) { | |
133 | pr_err("md-cluster: Unable to allocate resource name for resource %s\n", name); | |
134 | goto out_err; | |
135 | } | |
136 | strlcpy(res->name, name, namelen + 1); | |
137 | if (with_lvb) { | |
138 | res->lksb.sb_lvbptr = kzalloc(LVB_SIZE, GFP_KERNEL); | |
139 | if (!res->lksb.sb_lvbptr) { | |
140 | pr_err("md-cluster: Unable to allocate LVB for resource %s\n", name); | |
141 | goto out_err; | |
142 | } | |
143 | res->flags = DLM_LKF_VALBLK; | |
144 | } | |
145 | ||
146 | if (bastfn) | |
147 | res->bast = bastfn; | |
148 | ||
149 | res->flags |= DLM_LKF_EXPEDITE; | |
150 | ||
151 | ret = dlm_lock_sync(res, DLM_LOCK_NL); | |
152 | if (ret) { | |
153 | pr_err("md-cluster: Unable to lock NL on new lock resource %s\n", name); | |
154 | goto out_err; | |
155 | } | |
156 | res->flags &= ~DLM_LKF_EXPEDITE; | |
157 | res->flags |= DLM_LKF_CONVERT; | |
158 | ||
159 | return res; | |
160 | out_err: | |
161 | kfree(res->lksb.sb_lvbptr); | |
162 | kfree(res->name); | |
163 | kfree(res); | |
164 | return NULL; | |
165 | } | |
166 | ||
167 | static void lockres_free(struct dlm_lock_resource *res) | |
168 | { | |
b5ef5678 GJ |
169 | int ret; |
170 | ||
47741b7c GR |
171 | if (!res) |
172 | return; | |
173 | ||
b5ef5678 GJ |
174 | /* cancel a lock request or a conversion request that is blocked */ |
175 | res->flags |= DLM_LKF_CANCEL; | |
176 | retry: | |
177 | ret = dlm_unlock(res->ls, res->lksb.sb_lkid, 0, &res->lksb, res); | |
178 | if (unlikely(ret != 0)) { | |
179 | pr_info("%s: failed to unlock %s return %d\n", __func__, res->name, ret); | |
180 | ||
181 | /* if a lock conversion is cancelled, then the lock is put | |
182 | * back to grant queue, need to ensure it is unlocked */ | |
183 | if (ret == -DLM_ECANCEL) | |
184 | goto retry; | |
185 | } | |
186 | res->flags &= ~DLM_LKF_CANCEL; | |
47741b7c GR |
187 | wait_for_completion(&res->completion); |
188 | ||
189 | kfree(res->name); | |
190 | kfree(res->lksb.sb_lvbptr); | |
191 | kfree(res); | |
192 | } | |
8e854e9c | 193 | |
96ae923a GR |
194 | static void add_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres, |
195 | sector_t lo, sector_t hi) | |
196 | { | |
197 | struct resync_info *ri; | |
198 | ||
199 | ri = (struct resync_info *)lockres->lksb.sb_lvbptr; | |
200 | ri->lo = cpu_to_le64(lo); | |
201 | ri->hi = cpu_to_le64(hi); | |
202 | } | |
203 | ||
204 | static struct suspend_info *read_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres) | |
205 | { | |
206 | struct resync_info ri; | |
207 | struct suspend_info *s = NULL; | |
208 | sector_t hi = 0; | |
209 | ||
210 | dlm_lock_sync(lockres, DLM_LOCK_CR); | |
211 | memcpy(&ri, lockres->lksb.sb_lvbptr, sizeof(struct resync_info)); | |
212 | hi = le64_to_cpu(ri.hi); | |
213 | if (ri.hi > 0) { | |
214 | s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL); | |
215 | if (!s) | |
216 | goto out; | |
217 | s->hi = hi; | |
218 | s->lo = le64_to_cpu(ri.lo); | |
219 | } | |
220 | dlm_unlock_sync(lockres); | |
221 | out: | |
222 | return s; | |
223 | } | |
224 | ||
6dc69c9c | 225 | static void recover_bitmaps(struct md_thread *thread) |
e94987db GR |
226 | { |
227 | struct mddev *mddev = thread->mddev; | |
228 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
229 | struct dlm_lock_resource *bm_lockres; | |
230 | char str[64]; | |
231 | int slot, ret; | |
232 | struct suspend_info *s, *tmp; | |
233 | sector_t lo, hi; | |
234 | ||
235 | while (cinfo->recovery_map) { | |
236 | slot = fls64((u64)cinfo->recovery_map) - 1; | |
237 | ||
238 | /* Clear suspend_area associated with the bitmap */ | |
239 | spin_lock_irq(&cinfo->suspend_lock); | |
240 | list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list) | |
241 | if (slot == s->slot) { | |
242 | list_del(&s->list); | |
243 | kfree(s); | |
244 | } | |
245 | spin_unlock_irq(&cinfo->suspend_lock); | |
246 | ||
247 | snprintf(str, 64, "bitmap%04d", slot); | |
248 | bm_lockres = lockres_init(mddev, str, NULL, 1); | |
249 | if (!bm_lockres) { | |
250 | pr_err("md-cluster: Cannot initialize bitmaps\n"); | |
251 | goto clear_bit; | |
252 | } | |
253 | ||
254 | ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW); | |
255 | if (ret) { | |
256 | pr_err("md-cluster: Could not DLM lock %s: %d\n", | |
257 | str, ret); | |
258 | goto clear_bit; | |
259 | } | |
97f6cd39 | 260 | ret = bitmap_copy_from_slot(mddev, slot, &lo, &hi, true); |
4b26a08a | 261 | if (ret) { |
e94987db | 262 | pr_err("md-cluster: Could not copy data from bitmap %d\n", slot); |
4b26a08a GR |
263 | goto dlm_unlock; |
264 | } | |
265 | if (hi > 0) { | |
266 | /* TODO:Wait for current resync to get over */ | |
267 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
268 | if (lo < mddev->recovery_cp) | |
269 | mddev->recovery_cp = lo; | |
270 | md_check_recovery(mddev); | |
271 | } | |
272 | dlm_unlock: | |
e94987db GR |
273 | dlm_unlock_sync(bm_lockres); |
274 | clear_bit: | |
275 | clear_bit(slot, &cinfo->recovery_map); | |
276 | } | |
277 | } | |
278 | ||
cf921cc1 GR |
279 | static void recover_prep(void *arg) |
280 | { | |
90382ed9 GR |
281 | struct mddev *mddev = arg; |
282 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
283 | set_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state); | |
cf921cc1 GR |
284 | } |
285 | ||
05cd0e51 | 286 | static void __recover_slot(struct mddev *mddev, int slot) |
cf921cc1 | 287 | { |
cf921cc1 GR |
288 | struct md_cluster_info *cinfo = mddev->cluster_info; |
289 | ||
05cd0e51 | 290 | set_bit(slot, &cinfo->recovery_map); |
e94987db GR |
291 | if (!cinfo->recovery_thread) { |
292 | cinfo->recovery_thread = md_register_thread(recover_bitmaps, | |
293 | mddev, "recover"); | |
294 | if (!cinfo->recovery_thread) { | |
295 | pr_warn("md-cluster: Could not create recovery thread\n"); | |
296 | return; | |
297 | } | |
298 | } | |
299 | md_wakeup_thread(cinfo->recovery_thread); | |
cf921cc1 GR |
300 | } |
301 | ||
05cd0e51 GJ |
302 | static void recover_slot(void *arg, struct dlm_slot *slot) |
303 | { | |
304 | struct mddev *mddev = arg; | |
305 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
306 | ||
307 | pr_info("md-cluster: %s Node %d/%d down. My slot: %d. Initiating recovery.\n", | |
308 | mddev->bitmap_info.cluster_name, | |
309 | slot->nodeid, slot->slot, | |
310 | cinfo->slot_number); | |
311 | /* deduct one since dlm slot starts from one while the num of | |
312 | * cluster-md begins with 0 */ | |
313 | __recover_slot(mddev, slot->slot - 1); | |
314 | } | |
315 | ||
cf921cc1 GR |
316 | static void recover_done(void *arg, struct dlm_slot *slots, |
317 | int num_slots, int our_slot, | |
318 | uint32_t generation) | |
319 | { | |
320 | struct mddev *mddev = arg; | |
321 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
322 | ||
323 | cinfo->slot_number = our_slot; | |
eece075c GJ |
324 | /* completion is only need to be complete when node join cluster, |
325 | * it doesn't need to run during another node's failure */ | |
326 | if (test_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state)) { | |
327 | complete(&cinfo->completion); | |
328 | clear_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state); | |
329 | } | |
90382ed9 | 330 | clear_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state); |
cf921cc1 GR |
331 | } |
332 | ||
eece075c GJ |
333 | /* the ops is called when node join the cluster, and do lock recovery |
334 | * if node failure occurs */ | |
cf921cc1 GR |
335 | static const struct dlm_lockspace_ops md_ls_ops = { |
336 | .recover_prep = recover_prep, | |
337 | .recover_slot = recover_slot, | |
338 | .recover_done = recover_done, | |
339 | }; | |
340 | ||
4664680c GR |
341 | /* |
342 | * The BAST function for the ack lock resource | |
343 | * This function wakes up the receive thread in | |
344 | * order to receive and process the message. | |
345 | */ | |
346 | static void ack_bast(void *arg, int mode) | |
347 | { | |
348 | struct dlm_lock_resource *res = (struct dlm_lock_resource *)arg; | |
349 | struct md_cluster_info *cinfo = res->mddev->cluster_info; | |
350 | ||
351 | if (mode == DLM_LOCK_EX) | |
352 | md_wakeup_thread(cinfo->recv_thread); | |
353 | } | |
354 | ||
e59721cc GR |
355 | static void __remove_suspend_info(struct md_cluster_info *cinfo, int slot) |
356 | { | |
357 | struct suspend_info *s, *tmp; | |
358 | ||
359 | list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list) | |
360 | if (slot == s->slot) { | |
361 | pr_info("%s:%d Deleting suspend_info: %d\n", | |
362 | __func__, __LINE__, slot); | |
363 | list_del(&s->list); | |
364 | kfree(s); | |
365 | break; | |
366 | } | |
367 | } | |
368 | ||
b8ca846e | 369 | static void remove_suspend_info(struct mddev *mddev, int slot) |
e59721cc | 370 | { |
b8ca846e | 371 | struct md_cluster_info *cinfo = mddev->cluster_info; |
e59721cc GR |
372 | spin_lock_irq(&cinfo->suspend_lock); |
373 | __remove_suspend_info(cinfo, slot); | |
374 | spin_unlock_irq(&cinfo->suspend_lock); | |
b8ca846e | 375 | mddev->pers->quiesce(mddev, 2); |
e59721cc GR |
376 | } |
377 | ||
378 | ||
9ed38ff5 | 379 | static void process_suspend_info(struct mddev *mddev, |
e59721cc GR |
380 | int slot, sector_t lo, sector_t hi) |
381 | { | |
9ed38ff5 | 382 | struct md_cluster_info *cinfo = mddev->cluster_info; |
e59721cc GR |
383 | struct suspend_info *s; |
384 | ||
385 | if (!hi) { | |
b8ca846e | 386 | remove_suspend_info(mddev, slot); |
e59721cc GR |
387 | return; |
388 | } | |
389 | s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL); | |
390 | if (!s) | |
391 | return; | |
392 | s->slot = slot; | |
393 | s->lo = lo; | |
394 | s->hi = hi; | |
9ed38ff5 GR |
395 | mddev->pers->quiesce(mddev, 1); |
396 | mddev->pers->quiesce(mddev, 0); | |
e59721cc GR |
397 | spin_lock_irq(&cinfo->suspend_lock); |
398 | /* Remove existing entry (if exists) before adding */ | |
399 | __remove_suspend_info(cinfo, slot); | |
400 | list_add(&s->list, &cinfo->suspend_list); | |
401 | spin_unlock_irq(&cinfo->suspend_lock); | |
b8ca846e | 402 | mddev->pers->quiesce(mddev, 2); |
e59721cc GR |
403 | } |
404 | ||
1aee41f6 GR |
405 | static void process_add_new_disk(struct mddev *mddev, struct cluster_msg *cmsg) |
406 | { | |
407 | char disk_uuid[64]; | |
408 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
409 | char event_name[] = "EVENT=ADD_DEVICE"; | |
410 | char raid_slot[16]; | |
411 | char *envp[] = {event_name, disk_uuid, raid_slot, NULL}; | |
412 | int len; | |
413 | ||
414 | len = snprintf(disk_uuid, 64, "DEVICE_UUID="); | |
b89f704a | 415 | sprintf(disk_uuid + len, "%pU", cmsg->uuid); |
1aee41f6 GR |
416 | snprintf(raid_slot, 16, "RAID_DISK=%d", cmsg->raid_slot); |
417 | pr_info("%s:%d Sending kobject change with %s and %s\n", __func__, __LINE__, disk_uuid, raid_slot); | |
418 | init_completion(&cinfo->newdisk_completion); | |
fa8259da | 419 | set_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state); |
1aee41f6 GR |
420 | kobject_uevent_env(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE, envp); |
421 | wait_for_completion_timeout(&cinfo->newdisk_completion, | |
422 | NEW_DEV_TIMEOUT); | |
fa8259da | 423 | clear_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state); |
1aee41f6 GR |
424 | } |
425 | ||
426 | ||
427 | static void process_metadata_update(struct mddev *mddev, struct cluster_msg *msg) | |
428 | { | |
429 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
70bcecdb | 430 | md_reload_sb(mddev, le32_to_cpu(msg->raid_slot)); |
1aee41f6 GR |
431 | dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR); |
432 | } | |
433 | ||
88bcfef7 GR |
434 | static void process_remove_disk(struct mddev *mddev, struct cluster_msg *msg) |
435 | { | |
436 | struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev, msg->raid_slot); | |
437 | ||
438 | if (rdev) | |
439 | md_kick_rdev_from_array(rdev); | |
440 | else | |
441 | pr_warn("%s: %d Could not find disk(%d) to REMOVE\n", __func__, __LINE__, msg->raid_slot); | |
442 | } | |
443 | ||
97f6cd39 GR |
444 | static void process_readd_disk(struct mddev *mddev, struct cluster_msg *msg) |
445 | { | |
446 | struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev, msg->raid_slot); | |
447 | ||
448 | if (rdev && test_bit(Faulty, &rdev->flags)) | |
449 | clear_bit(Faulty, &rdev->flags); | |
450 | else | |
451 | pr_warn("%s: %d Could not find disk(%d) which is faulty", __func__, __LINE__, msg->raid_slot); | |
452 | } | |
453 | ||
4664680c GR |
454 | static void process_recvd_msg(struct mddev *mddev, struct cluster_msg *msg) |
455 | { | |
456 | switch (msg->type) { | |
457 | case METADATA_UPDATED: | |
458 | pr_info("%s: %d Received message: METADATA_UPDATE from %d\n", | |
459 | __func__, __LINE__, msg->slot); | |
1aee41f6 | 460 | process_metadata_update(mddev, msg); |
4664680c GR |
461 | break; |
462 | case RESYNCING: | |
463 | pr_info("%s: %d Received message: RESYNCING from %d\n", | |
464 | __func__, __LINE__, msg->slot); | |
9ed38ff5 | 465 | process_suspend_info(mddev, msg->slot, |
e59721cc | 466 | msg->low, msg->high); |
4664680c | 467 | break; |
1aee41f6 GR |
468 | case NEWDISK: |
469 | pr_info("%s: %d Received message: NEWDISK from %d\n", | |
470 | __func__, __LINE__, msg->slot); | |
471 | process_add_new_disk(mddev, msg); | |
88bcfef7 GR |
472 | break; |
473 | case REMOVE: | |
474 | pr_info("%s: %d Received REMOVE from %d\n", | |
475 | __func__, __LINE__, msg->slot); | |
476 | process_remove_disk(mddev, msg); | |
477 | break; | |
97f6cd39 GR |
478 | case RE_ADD: |
479 | pr_info("%s: %d Received RE_ADD from %d\n", | |
480 | __func__, __LINE__, msg->slot); | |
481 | process_readd_disk(mddev, msg); | |
482 | break; | |
dc737d7c GJ |
483 | case BITMAP_NEEDS_SYNC: |
484 | pr_info("%s: %d Received BITMAP_NEEDS_SYNC from %d\n", | |
485 | __func__, __LINE__, msg->slot); | |
486 | __recover_slot(mddev, msg->slot); | |
487 | break; | |
88bcfef7 GR |
488 | default: |
489 | pr_warn("%s:%d Received unknown message from %d\n", | |
490 | __func__, __LINE__, msg->slot); | |
09dd1af2 | 491 | } |
4664680c GR |
492 | } |
493 | ||
494 | /* | |
495 | * thread for receiving message | |
496 | */ | |
497 | static void recv_daemon(struct md_thread *thread) | |
498 | { | |
499 | struct md_cluster_info *cinfo = thread->mddev->cluster_info; | |
500 | struct dlm_lock_resource *ack_lockres = cinfo->ack_lockres; | |
501 | struct dlm_lock_resource *message_lockres = cinfo->message_lockres; | |
502 | struct cluster_msg msg; | |
b5ef5678 | 503 | int ret; |
4664680c GR |
504 | |
505 | /*get CR on Message*/ | |
506 | if (dlm_lock_sync(message_lockres, DLM_LOCK_CR)) { | |
507 | pr_err("md/raid1:failed to get CR on MESSAGE\n"); | |
508 | return; | |
509 | } | |
510 | ||
511 | /* read lvb and wake up thread to process this message_lockres */ | |
512 | memcpy(&msg, message_lockres->lksb.sb_lvbptr, sizeof(struct cluster_msg)); | |
513 | process_recvd_msg(thread->mddev, &msg); | |
514 | ||
515 | /*release CR on ack_lockres*/ | |
b5ef5678 GJ |
516 | ret = dlm_unlock_sync(ack_lockres); |
517 | if (unlikely(ret != 0)) | |
518 | pr_info("unlock ack failed return %d\n", ret); | |
66099bb0 | 519 | /*up-convert to PR on message_lockres*/ |
b5ef5678 GJ |
520 | ret = dlm_lock_sync(message_lockres, DLM_LOCK_PR); |
521 | if (unlikely(ret != 0)) | |
522 | pr_info("lock PR on msg failed return %d\n", ret); | |
4664680c | 523 | /*get CR on ack_lockres again*/ |
b5ef5678 GJ |
524 | ret = dlm_lock_sync(ack_lockres, DLM_LOCK_CR); |
525 | if (unlikely(ret != 0)) | |
526 | pr_info("lock CR on ack failed return %d\n", ret); | |
4664680c | 527 | /*release CR on message_lockres*/ |
b5ef5678 GJ |
528 | ret = dlm_unlock_sync(message_lockres); |
529 | if (unlikely(ret != 0)) | |
530 | pr_info("unlock msg failed return %d\n", ret); | |
4664680c GR |
531 | } |
532 | ||
601b515c GR |
533 | /* lock_comm() |
534 | * Takes the lock on the TOKEN lock resource so no other | |
535 | * node can communicate while the operation is underway. | |
536 | */ | |
537 | static int lock_comm(struct md_cluster_info *cinfo) | |
538 | { | |
539 | int error; | |
540 | ||
541 | error = dlm_lock_sync(cinfo->token_lockres, DLM_LOCK_EX); | |
542 | if (error) | |
543 | pr_err("md-cluster(%s:%d): failed to get EX on TOKEN (%d)\n", | |
544 | __func__, __LINE__, error); | |
545 | return error; | |
546 | } | |
547 | ||
548 | static void unlock_comm(struct md_cluster_info *cinfo) | |
549 | { | |
550 | dlm_unlock_sync(cinfo->token_lockres); | |
551 | } | |
552 | ||
553 | /* __sendmsg() | |
554 | * This function performs the actual sending of the message. This function is | |
555 | * usually called after performing the encompassing operation | |
556 | * The function: | |
557 | * 1. Grabs the message lockresource in EX mode | |
558 | * 2. Copies the message to the message LVB | |
66099bb0 | 559 | * 3. Downconverts message lockresource to CW |
601b515c GR |
560 | * 4. Upconverts ack lock resource from CR to EX. This forces the BAST on other nodes |
561 | * and the other nodes read the message. The thread will wait here until all other | |
562 | * nodes have released ack lock resource. | |
563 | * 5. Downconvert ack lockresource to CR | |
564 | */ | |
565 | static int __sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg) | |
566 | { | |
567 | int error; | |
568 | int slot = cinfo->slot_number - 1; | |
569 | ||
570 | cmsg->slot = cpu_to_le32(slot); | |
571 | /*get EX on Message*/ | |
572 | error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_EX); | |
573 | if (error) { | |
574 | pr_err("md-cluster: failed to get EX on MESSAGE (%d)\n", error); | |
575 | goto failed_message; | |
576 | } | |
577 | ||
578 | memcpy(cinfo->message_lockres->lksb.sb_lvbptr, (void *)cmsg, | |
579 | sizeof(struct cluster_msg)); | |
66099bb0 GJ |
580 | /*down-convert EX to CW on Message*/ |
581 | error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_CW); | |
601b515c | 582 | if (error) { |
66099bb0 | 583 | pr_err("md-cluster: failed to convert EX to CW on MESSAGE(%d)\n", |
601b515c | 584 | error); |
66099bb0 | 585 | goto failed_ack; |
601b515c GR |
586 | } |
587 | ||
588 | /*up-convert CR to EX on Ack*/ | |
589 | error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_EX); | |
590 | if (error) { | |
591 | pr_err("md-cluster: failed to convert CR to EX on ACK(%d)\n", | |
592 | error); | |
593 | goto failed_ack; | |
594 | } | |
595 | ||
596 | /*down-convert EX to CR on Ack*/ | |
597 | error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR); | |
598 | if (error) { | |
599 | pr_err("md-cluster: failed to convert EX to CR on ACK(%d)\n", | |
600 | error); | |
601 | goto failed_ack; | |
602 | } | |
603 | ||
604 | failed_ack: | |
b5ef5678 GJ |
605 | error = dlm_unlock_sync(cinfo->message_lockres); |
606 | if (unlikely(error != 0)) { | |
607 | pr_err("md-cluster: failed convert to NL on MESSAGE(%d)\n", | |
608 | error); | |
609 | /* in case the message can't be released due to some reason */ | |
610 | goto failed_ack; | |
611 | } | |
601b515c GR |
612 | failed_message: |
613 | return error; | |
614 | } | |
615 | ||
616 | static int sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg) | |
617 | { | |
618 | int ret; | |
619 | ||
620 | lock_comm(cinfo); | |
621 | ret = __sendmsg(cinfo, cmsg); | |
622 | unlock_comm(cinfo); | |
623 | return ret; | |
624 | } | |
625 | ||
96ae923a GR |
626 | static int gather_all_resync_info(struct mddev *mddev, int total_slots) |
627 | { | |
628 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
629 | int i, ret = 0; | |
630 | struct dlm_lock_resource *bm_lockres; | |
631 | struct suspend_info *s; | |
632 | char str[64]; | |
abb9b22a | 633 | sector_t lo, hi; |
96ae923a GR |
634 | |
635 | ||
636 | for (i = 0; i < total_slots; i++) { | |
637 | memset(str, '\0', 64); | |
638 | snprintf(str, 64, "bitmap%04d", i); | |
639 | bm_lockres = lockres_init(mddev, str, NULL, 1); | |
640 | if (!bm_lockres) | |
641 | return -ENOMEM; | |
642 | if (i == (cinfo->slot_number - 1)) | |
643 | continue; | |
644 | ||
645 | bm_lockres->flags |= DLM_LKF_NOQUEUE; | |
646 | ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW); | |
647 | if (ret == -EAGAIN) { | |
648 | memset(bm_lockres->lksb.sb_lvbptr, '\0', LVB_SIZE); | |
649 | s = read_resync_info(mddev, bm_lockres); | |
650 | if (s) { | |
651 | pr_info("%s:%d Resync[%llu..%llu] in progress on %d\n", | |
652 | __func__, __LINE__, | |
653 | (unsigned long long) s->lo, | |
654 | (unsigned long long) s->hi, i); | |
655 | spin_lock_irq(&cinfo->suspend_lock); | |
656 | s->slot = i; | |
657 | list_add(&s->list, &cinfo->suspend_list); | |
658 | spin_unlock_irq(&cinfo->suspend_lock); | |
659 | } | |
660 | ret = 0; | |
661 | lockres_free(bm_lockres); | |
662 | continue; | |
663 | } | |
6e6d9f2c GJ |
664 | if (ret) { |
665 | lockres_free(bm_lockres); | |
96ae923a | 666 | goto out; |
6e6d9f2c | 667 | } |
abb9b22a GJ |
668 | |
669 | /* Read the disk bitmap sb and check if it needs recovery */ | |
670 | ret = bitmap_copy_from_slot(mddev, i, &lo, &hi, false); | |
671 | if (ret) { | |
672 | pr_warn("md-cluster: Could not gather bitmaps from slot %d", i); | |
673 | lockres_free(bm_lockres); | |
674 | continue; | |
675 | } | |
676 | if ((hi > 0) && (lo < mddev->recovery_cp)) { | |
677 | set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); | |
678 | mddev->recovery_cp = lo; | |
679 | md_check_recovery(mddev); | |
680 | } | |
681 | ||
96ae923a GR |
682 | dlm_unlock_sync(bm_lockres); |
683 | lockres_free(bm_lockres); | |
684 | } | |
685 | out: | |
686 | return ret; | |
687 | } | |
688 | ||
edb39c9d GR |
689 | static int join(struct mddev *mddev, int nodes) |
690 | { | |
c4ce867f | 691 | struct md_cluster_info *cinfo; |
cf921cc1 | 692 | int ret, ops_rv; |
c4ce867f GR |
693 | char str[64]; |
694 | ||
c4ce867f GR |
695 | cinfo = kzalloc(sizeof(struct md_cluster_info), GFP_KERNEL); |
696 | if (!cinfo) | |
697 | return -ENOMEM; | |
698 | ||
9e3072e3 GJ |
699 | INIT_LIST_HEAD(&cinfo->suspend_list); |
700 | spin_lock_init(&cinfo->suspend_lock); | |
cf921cc1 | 701 | init_completion(&cinfo->completion); |
eece075c | 702 | set_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state); |
cf921cc1 GR |
703 | |
704 | mutex_init(&cinfo->sb_mutex); | |
705 | mddev->cluster_info = cinfo; | |
706 | ||
c4ce867f | 707 | memset(str, 0, 64); |
b89f704a | 708 | sprintf(str, "%pU", mddev->uuid); |
cf921cc1 GR |
709 | ret = dlm_new_lockspace(str, mddev->bitmap_info.cluster_name, |
710 | DLM_LSFL_FS, LVB_SIZE, | |
711 | &md_ls_ops, mddev, &ops_rv, &cinfo->lockspace); | |
c4ce867f GR |
712 | if (ret) |
713 | goto err; | |
cf921cc1 | 714 | wait_for_completion(&cinfo->completion); |
8c58f02e GJ |
715 | if (nodes < cinfo->slot_number) { |
716 | pr_err("md-cluster: Slot allotted(%d) is greater than available slots(%d).", | |
717 | cinfo->slot_number, nodes); | |
b97e9257 GR |
718 | ret = -ERANGE; |
719 | goto err; | |
720 | } | |
4664680c GR |
721 | /* Initiate the communication resources */ |
722 | ret = -ENOMEM; | |
723 | cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv"); | |
724 | if (!cinfo->recv_thread) { | |
725 | pr_err("md-cluster: cannot allocate memory for recv_thread!\n"); | |
726 | goto err; | |
727 | } | |
728 | cinfo->message_lockres = lockres_init(mddev, "message", NULL, 1); | |
729 | if (!cinfo->message_lockres) | |
730 | goto err; | |
731 | cinfo->token_lockres = lockres_init(mddev, "token", NULL, 0); | |
732 | if (!cinfo->token_lockres) | |
733 | goto err; | |
734 | cinfo->ack_lockres = lockres_init(mddev, "ack", ack_bast, 0); | |
735 | if (!cinfo->ack_lockres) | |
736 | goto err; | |
1aee41f6 GR |
737 | cinfo->no_new_dev_lockres = lockres_init(mddev, "no-new-dev", NULL, 0); |
738 | if (!cinfo->no_new_dev_lockres) | |
739 | goto err; | |
740 | ||
4664680c GR |
741 | /* get sync CR lock on ACK. */ |
742 | if (dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR)) | |
743 | pr_err("md-cluster: failed to get a sync CR lock on ACK!(%d)\n", | |
744 | ret); | |
1aee41f6 GR |
745 | /* get sync CR lock on no-new-dev. */ |
746 | if (dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR)) | |
747 | pr_err("md-cluster: failed to get a sync CR lock on no-new-dev!(%d)\n", ret); | |
748 | ||
54519c5f GR |
749 | |
750 | pr_info("md-cluster: Joined cluster %s slot %d\n", str, cinfo->slot_number); | |
751 | snprintf(str, 64, "bitmap%04d", cinfo->slot_number - 1); | |
752 | cinfo->bitmap_lockres = lockres_init(mddev, str, NULL, 1); | |
753 | if (!cinfo->bitmap_lockres) | |
754 | goto err; | |
755 | if (dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW)) { | |
756 | pr_err("Failed to get bitmap lock\n"); | |
757 | ret = -EINVAL; | |
758 | goto err; | |
759 | } | |
760 | ||
96ae923a GR |
761 | ret = gather_all_resync_info(mddev, nodes); |
762 | if (ret) | |
763 | goto err; | |
764 | ||
edb39c9d | 765 | return 0; |
c4ce867f | 766 | err: |
4664680c GR |
767 | lockres_free(cinfo->message_lockres); |
768 | lockres_free(cinfo->token_lockres); | |
769 | lockres_free(cinfo->ack_lockres); | |
1aee41f6 | 770 | lockres_free(cinfo->no_new_dev_lockres); |
96ae923a | 771 | lockres_free(cinfo->bitmap_lockres); |
c4ce867f GR |
772 | if (cinfo->lockspace) |
773 | dlm_release_lockspace(cinfo->lockspace, 2); | |
cf921cc1 | 774 | mddev->cluster_info = NULL; |
c4ce867f | 775 | kfree(cinfo); |
c4ce867f | 776 | return ret; |
edb39c9d GR |
777 | } |
778 | ||
09995411 GJ |
779 | static void resync_bitmap(struct mddev *mddev) |
780 | { | |
781 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
782 | struct cluster_msg cmsg = {0}; | |
783 | int err; | |
784 | ||
785 | cmsg.type = cpu_to_le32(BITMAP_NEEDS_SYNC); | |
786 | err = sendmsg(cinfo, &cmsg); | |
787 | if (err) | |
788 | pr_err("%s:%d: failed to send BITMAP_NEEDS_SYNC message (%d)\n", | |
789 | __func__, __LINE__, err); | |
790 | } | |
791 | ||
edb39c9d GR |
792 | static int leave(struct mddev *mddev) |
793 | { | |
c4ce867f GR |
794 | struct md_cluster_info *cinfo = mddev->cluster_info; |
795 | ||
796 | if (!cinfo) | |
797 | return 0; | |
09995411 GJ |
798 | |
799 | /* BITMAP_NEEDS_SYNC message should be sent when node | |
800 | * is leaving the cluster with dirty bitmap, also we | |
801 | * can only deliver it when dlm connection is available */ | |
802 | if (cinfo->slot_number > 0 && mddev->recovery_cp != MaxSector) | |
803 | resync_bitmap(mddev); | |
804 | ||
e94987db | 805 | md_unregister_thread(&cinfo->recovery_thread); |
4664680c GR |
806 | md_unregister_thread(&cinfo->recv_thread); |
807 | lockres_free(cinfo->message_lockres); | |
808 | lockres_free(cinfo->token_lockres); | |
809 | lockres_free(cinfo->ack_lockres); | |
1aee41f6 | 810 | lockres_free(cinfo->no_new_dev_lockres); |
54519c5f | 811 | lockres_free(cinfo->bitmap_lockres); |
c4ce867f | 812 | dlm_release_lockspace(cinfo->lockspace, 2); |
edb39c9d GR |
813 | return 0; |
814 | } | |
815 | ||
cf921cc1 GR |
816 | /* slot_number(): Returns the MD slot number to use |
817 | * DLM starts the slot numbers from 1, wheras cluster-md | |
818 | * wants the number to be from zero, so we deduct one | |
819 | */ | |
820 | static int slot_number(struct mddev *mddev) | |
821 | { | |
822 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
823 | ||
824 | return cinfo->slot_number - 1; | |
825 | } | |
826 | ||
293467aa GR |
827 | static int metadata_update_start(struct mddev *mddev) |
828 | { | |
829 | return lock_comm(mddev->cluster_info); | |
830 | } | |
831 | ||
832 | static int metadata_update_finish(struct mddev *mddev) | |
833 | { | |
834 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
835 | struct cluster_msg cmsg; | |
70bcecdb GR |
836 | struct md_rdev *rdev; |
837 | int ret = 0; | |
293467aa GR |
838 | |
839 | memset(&cmsg, 0, sizeof(cmsg)); | |
840 | cmsg.type = cpu_to_le32(METADATA_UPDATED); | |
70bcecdb GR |
841 | cmsg.raid_slot = -1; |
842 | /* Pick up a good active device number to send. | |
843 | */ | |
844 | rdev_for_each(rdev, mddev) | |
845 | if (rdev->raid_disk > -1 && !test_bit(Faulty, &rdev->flags)) { | |
846 | cmsg.raid_slot = cpu_to_le32(rdev->desc_nr); | |
847 | break; | |
848 | } | |
849 | if (cmsg.raid_slot >= 0) | |
850 | ret = __sendmsg(cinfo, &cmsg); | |
851 | else | |
852 | pr_warn("md-cluster: No good device id found to send\n"); | |
293467aa GR |
853 | unlock_comm(cinfo); |
854 | return ret; | |
855 | } | |
856 | ||
857 | static int metadata_update_cancel(struct mddev *mddev) | |
858 | { | |
859 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
860 | ||
861 | return dlm_unlock_sync(cinfo->token_lockres); | |
862 | } | |
863 | ||
c40f341f | 864 | static int resync_info_update(struct mddev *mddev, sector_t lo, sector_t hi) |
965400eb GR |
865 | { |
866 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
867 | struct cluster_msg cmsg; | |
868 | int slot = cinfo->slot_number - 1; | |
869 | ||
c40f341f GR |
870 | add_resync_info(mddev, cinfo->bitmap_lockres, lo, hi); |
871 | /* Re-acquire the lock to refresh LVB */ | |
872 | dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW); | |
965400eb GR |
873 | pr_info("%s:%d lo: %llu hi: %llu\n", __func__, __LINE__, |
874 | (unsigned long long)lo, | |
875 | (unsigned long long)hi); | |
c40f341f | 876 | cmsg.type = cpu_to_le32(RESYNCING); |
965400eb GR |
877 | cmsg.slot = cpu_to_le32(slot); |
878 | cmsg.low = cpu_to_le64(lo); | |
879 | cmsg.high = cpu_to_le64(hi); | |
880 | return sendmsg(cinfo, &cmsg); | |
881 | } | |
882 | ||
90382ed9 GR |
883 | static int area_resyncing(struct mddev *mddev, int direction, |
884 | sector_t lo, sector_t hi) | |
589a1c49 GR |
885 | { |
886 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
887 | int ret = 0; | |
888 | struct suspend_info *s; | |
889 | ||
90382ed9 GR |
890 | if ((direction == READ) && |
891 | test_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state)) | |
892 | return 1; | |
893 | ||
589a1c49 GR |
894 | spin_lock_irq(&cinfo->suspend_lock); |
895 | if (list_empty(&cinfo->suspend_list)) | |
896 | goto out; | |
897 | list_for_each_entry(s, &cinfo->suspend_list, list) | |
898 | if (hi > s->lo && lo < s->hi) { | |
899 | ret = 1; | |
900 | break; | |
901 | } | |
902 | out: | |
903 | spin_unlock_irq(&cinfo->suspend_lock); | |
904 | return ret; | |
905 | } | |
906 | ||
1aee41f6 GR |
907 | static int add_new_disk_start(struct mddev *mddev, struct md_rdev *rdev) |
908 | { | |
909 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
910 | struct cluster_msg cmsg; | |
911 | int ret = 0; | |
912 | struct mdp_superblock_1 *sb = page_address(rdev->sb_page); | |
913 | char *uuid = sb->device_uuid; | |
914 | ||
915 | memset(&cmsg, 0, sizeof(cmsg)); | |
916 | cmsg.type = cpu_to_le32(NEWDISK); | |
917 | memcpy(cmsg.uuid, uuid, 16); | |
918 | cmsg.raid_slot = rdev->desc_nr; | |
919 | lock_comm(cinfo); | |
920 | ret = __sendmsg(cinfo, &cmsg); | |
921 | if (ret) | |
922 | return ret; | |
923 | cinfo->no_new_dev_lockres->flags |= DLM_LKF_NOQUEUE; | |
924 | ret = dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_EX); | |
925 | cinfo->no_new_dev_lockres->flags &= ~DLM_LKF_NOQUEUE; | |
926 | /* Some node does not "see" the device */ | |
927 | if (ret == -EAGAIN) | |
928 | ret = -ENOENT; | |
929 | else | |
930 | dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR); | |
931 | return ret; | |
932 | } | |
933 | ||
934 | static int add_new_disk_finish(struct mddev *mddev) | |
935 | { | |
1aee41f6 GR |
936 | /* Write sb and inform others */ |
937 | md_update_sb(mddev, 1); | |
70bcecdb | 938 | return metadata_update_finish(mddev); |
1aee41f6 GR |
939 | } |
940 | ||
fa8259da | 941 | static int new_disk_ack(struct mddev *mddev, bool ack) |
1aee41f6 GR |
942 | { |
943 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
944 | ||
fa8259da GR |
945 | if (!test_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state)) { |
946 | pr_warn("md-cluster(%s): Spurious cluster confirmation\n", mdname(mddev)); | |
947 | return -EINVAL; | |
948 | } | |
949 | ||
1aee41f6 GR |
950 | if (ack) |
951 | dlm_unlock_sync(cinfo->no_new_dev_lockres); | |
952 | complete(&cinfo->newdisk_completion); | |
fa8259da | 953 | return 0; |
1aee41f6 GR |
954 | } |
955 | ||
88bcfef7 GR |
956 | static int remove_disk(struct mddev *mddev, struct md_rdev *rdev) |
957 | { | |
958 | struct cluster_msg cmsg; | |
959 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
960 | cmsg.type = REMOVE; | |
961 | cmsg.raid_slot = rdev->desc_nr; | |
962 | return __sendmsg(cinfo, &cmsg); | |
963 | } | |
964 | ||
97f6cd39 GR |
965 | static int gather_bitmaps(struct md_rdev *rdev) |
966 | { | |
967 | int sn, err; | |
968 | sector_t lo, hi; | |
969 | struct cluster_msg cmsg; | |
970 | struct mddev *mddev = rdev->mddev; | |
971 | struct md_cluster_info *cinfo = mddev->cluster_info; | |
972 | ||
973 | cmsg.type = RE_ADD; | |
974 | cmsg.raid_slot = rdev->desc_nr; | |
975 | err = sendmsg(cinfo, &cmsg); | |
976 | if (err) | |
977 | goto out; | |
978 | ||
979 | for (sn = 0; sn < mddev->bitmap_info.nodes; sn++) { | |
980 | if (sn == (cinfo->slot_number - 1)) | |
981 | continue; | |
982 | err = bitmap_copy_from_slot(mddev, sn, &lo, &hi, false); | |
983 | if (err) { | |
984 | pr_warn("md-cluster: Could not gather bitmaps from slot %d", sn); | |
985 | goto out; | |
986 | } | |
987 | if ((hi > 0) && (lo < mddev->recovery_cp)) | |
988 | mddev->recovery_cp = lo; | |
989 | } | |
990 | out: | |
991 | return err; | |
992 | } | |
993 | ||
edb39c9d GR |
994 | static struct md_cluster_operations cluster_ops = { |
995 | .join = join, | |
996 | .leave = leave, | |
cf921cc1 | 997 | .slot_number = slot_number, |
96ae923a | 998 | .resync_info_update = resync_info_update, |
293467aa GR |
999 | .metadata_update_start = metadata_update_start, |
1000 | .metadata_update_finish = metadata_update_finish, | |
1001 | .metadata_update_cancel = metadata_update_cancel, | |
589a1c49 | 1002 | .area_resyncing = area_resyncing, |
1aee41f6 GR |
1003 | .add_new_disk_start = add_new_disk_start, |
1004 | .add_new_disk_finish = add_new_disk_finish, | |
1005 | .new_disk_ack = new_disk_ack, | |
88bcfef7 | 1006 | .remove_disk = remove_disk, |
97f6cd39 | 1007 | .gather_bitmaps = gather_bitmaps, |
edb39c9d GR |
1008 | }; |
1009 | ||
8e854e9c GR |
1010 | static int __init cluster_init(void) |
1011 | { | |
1012 | pr_warn("md-cluster: EXPERIMENTAL. Use with caution\n"); | |
1013 | pr_info("Registering Cluster MD functions\n"); | |
edb39c9d | 1014 | register_md_cluster_operations(&cluster_ops, THIS_MODULE); |
8e854e9c GR |
1015 | return 0; |
1016 | } | |
1017 | ||
1018 | static void cluster_exit(void) | |
1019 | { | |
edb39c9d | 1020 | unregister_md_cluster_operations(); |
8e854e9c GR |
1021 | } |
1022 | ||
1023 | module_init(cluster_init); | |
1024 | module_exit(cluster_exit); | |
1025 | MODULE_LICENSE("GPL"); | |
1026 | MODULE_DESCRIPTION("Clustering support for MD"); |