xref: /linux/drivers/md/md-cluster.c (revision c0e297dc61f8d4453e07afbea1fa8d0e67cd4a34)
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>
13 #include <linux/dlm.h>
14 #include <linux/sched.h>
15 #include <linux/raid/md_p.h>
16 #include "md.h"
17 #include "bitmap.h"
18 #include "md-cluster.h"
19 
20 #define LVB_SIZE	64
21 #define NEW_DEV_TIMEOUT 5000
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() */
28 	struct completion completion; /* completion for synchronized locking */
29 	void (*bast)(void *arg, int mode); /* blocking AST function pointer*/
30 	struct mddev *mddev; /* pointing back to mddev. */
31 };
32 
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 
45 /* md_cluster_info flags */
46 #define		MD_CLUSTER_WAITING_FOR_NEWDISK		1
47 #define		MD_CLUSTER_SUSPEND_READ_BALANCING	2
48 
49 
50 struct md_cluster_info {
51 	/* dlm lock space and resources for clustered raid. */
52 	dlm_lockspace_t *lockspace;
53 	int slot_number;
54 	struct completion completion;
55 	struct dlm_lock_resource *sb_lock;
56 	struct mutex sb_mutex;
57 	struct dlm_lock_resource *bitmap_lockres;
58 	struct list_head suspend_list;
59 	spinlock_t suspend_lock;
60 	struct md_thread *recovery_thread;
61 	unsigned long recovery_map;
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;
66 	struct dlm_lock_resource *no_new_dev_lockres;
67 	struct md_thread *recv_thread;
68 	struct completion newdisk_completion;
69 	unsigned long state;
70 };
71 
72 enum msg_type {
73 	METADATA_UPDATED = 0,
74 	RESYNCING,
75 	NEWDISK,
76 	REMOVE,
77 	RE_ADD,
78 };
79 
80 struct cluster_msg {
81 	int type;
82 	int slot;
83 	/* TODO: Unionize this for smaller footprint */
84 	sector_t low;
85 	sector_t high;
86 	char uuid[16];
87 	int raid_slot;
88 };
89 
90 static void sync_ast(void *arg)
91 {
92 	struct dlm_lock_resource *res;
93 
94 	res = (struct dlm_lock_resource *) arg;
95 	complete(&res->completion);
96 }
97 
98 static int dlm_lock_sync(struct dlm_lock_resource *res, int mode)
99 {
100 	int ret = 0;
101 
102 	init_completion(&res->completion);
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 
117 static struct dlm_lock_resource *lockres_init(struct mddev *mddev,
118 		char *name, void (*bastfn)(void *arg, int mode), int with_lvb)
119 {
120 	struct dlm_lock_resource *res = NULL;
121 	int ret, namelen;
122 	struct md_cluster_info *cinfo = mddev->cluster_info;
123 
124 	res = kzalloc(sizeof(struct dlm_lock_resource), GFP_KERNEL);
125 	if (!res)
126 		return NULL;
127 	res->ls = cinfo->lockspace;
128 	res->mddev = mddev;
129 	namelen = strlen(name);
130 	res->name = kzalloc(namelen + 1, GFP_KERNEL);
131 	if (!res->name) {
132 		pr_err("md-cluster: Unable to allocate resource name for resource %s\n", name);
133 		goto out_err;
134 	}
135 	strlcpy(res->name, name, namelen + 1);
136 	if (with_lvb) {
137 		res->lksb.sb_lvbptr = kzalloc(LVB_SIZE, GFP_KERNEL);
138 		if (!res->lksb.sb_lvbptr) {
139 			pr_err("md-cluster: Unable to allocate LVB for resource %s\n", name);
140 			goto out_err;
141 		}
142 		res->flags = DLM_LKF_VALBLK;
143 	}
144 
145 	if (bastfn)
146 		res->bast = bastfn;
147 
148 	res->flags |= DLM_LKF_EXPEDITE;
149 
150 	ret = dlm_lock_sync(res, DLM_LOCK_NL);
151 	if (ret) {
152 		pr_err("md-cluster: Unable to lock NL on new lock resource %s\n", name);
153 		goto out_err;
154 	}
155 	res->flags &= ~DLM_LKF_EXPEDITE;
156 	res->flags |= DLM_LKF_CONVERT;
157 
158 	return res;
159 out_err:
160 	kfree(res->lksb.sb_lvbptr);
161 	kfree(res->name);
162 	kfree(res);
163 	return NULL;
164 }
165 
166 static void lockres_free(struct dlm_lock_resource *res)
167 {
168 	if (!res)
169 		return;
170 
171 	init_completion(&res->completion);
172 	dlm_unlock(res->ls, res->lksb.sb_lkid, 0, &res->lksb, res);
173 	wait_for_completion(&res->completion);
174 
175 	kfree(res->name);
176 	kfree(res->lksb.sb_lvbptr);
177 	kfree(res);
178 }
179 
180 static char *pretty_uuid(char *dest, char *src)
181 {
182 	int i, len = 0;
183 
184 	for (i = 0; i < 16; i++) {
185 		if (i == 4 || i == 6 || i == 8 || i == 10)
186 			len += sprintf(dest + len, "-");
187 		len += sprintf(dest + len, "%02x", (__u8)src[i]);
188 	}
189 	return dest;
190 }
191 
192 static void add_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres,
193 		sector_t lo, sector_t hi)
194 {
195 	struct resync_info *ri;
196 
197 	ri = (struct resync_info *)lockres->lksb.sb_lvbptr;
198 	ri->lo = cpu_to_le64(lo);
199 	ri->hi = cpu_to_le64(hi);
200 }
201 
202 static struct suspend_info *read_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres)
203 {
204 	struct resync_info ri;
205 	struct suspend_info *s = NULL;
206 	sector_t hi = 0;
207 
208 	dlm_lock_sync(lockres, DLM_LOCK_CR);
209 	memcpy(&ri, lockres->lksb.sb_lvbptr, sizeof(struct resync_info));
210 	hi = le64_to_cpu(ri.hi);
211 	if (ri.hi > 0) {
212 		s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL);
213 		if (!s)
214 			goto out;
215 		s->hi = hi;
216 		s->lo = le64_to_cpu(ri.lo);
217 	}
218 	dlm_unlock_sync(lockres);
219 out:
220 	return s;
221 }
222 
223 static void recover_bitmaps(struct md_thread *thread)
224 {
225 	struct mddev *mddev = thread->mddev;
226 	struct md_cluster_info *cinfo = mddev->cluster_info;
227 	struct dlm_lock_resource *bm_lockres;
228 	char str[64];
229 	int slot, ret;
230 	struct suspend_info *s, *tmp;
231 	sector_t lo, hi;
232 
233 	while (cinfo->recovery_map) {
234 		slot = fls64((u64)cinfo->recovery_map) - 1;
235 
236 		/* Clear suspend_area associated with the bitmap */
237 		spin_lock_irq(&cinfo->suspend_lock);
238 		list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list)
239 			if (slot == s->slot) {
240 				list_del(&s->list);
241 				kfree(s);
242 			}
243 		spin_unlock_irq(&cinfo->suspend_lock);
244 
245 		snprintf(str, 64, "bitmap%04d", slot);
246 		bm_lockres = lockres_init(mddev, str, NULL, 1);
247 		if (!bm_lockres) {
248 			pr_err("md-cluster: Cannot initialize bitmaps\n");
249 			goto clear_bit;
250 		}
251 
252 		ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
253 		if (ret) {
254 			pr_err("md-cluster: Could not DLM lock %s: %d\n",
255 					str, ret);
256 			goto clear_bit;
257 		}
258 		ret = bitmap_copy_from_slot(mddev, slot, &lo, &hi, true);
259 		if (ret) {
260 			pr_err("md-cluster: Could not copy data from bitmap %d\n", slot);
261 			goto dlm_unlock;
262 		}
263 		if (hi > 0) {
264 			/* TODO:Wait for current resync to get over */
265 			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
266 			if (lo < mddev->recovery_cp)
267 				mddev->recovery_cp = lo;
268 			md_check_recovery(mddev);
269 		}
270 dlm_unlock:
271 		dlm_unlock_sync(bm_lockres);
272 clear_bit:
273 		clear_bit(slot, &cinfo->recovery_map);
274 	}
275 }
276 
277 static void recover_prep(void *arg)
278 {
279 	struct mddev *mddev = arg;
280 	struct md_cluster_info *cinfo = mddev->cluster_info;
281 	set_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state);
282 }
283 
284 static void recover_slot(void *arg, struct dlm_slot *slot)
285 {
286 	struct mddev *mddev = arg;
287 	struct md_cluster_info *cinfo = mddev->cluster_info;
288 
289 	pr_info("md-cluster: %s Node %d/%d down. My slot: %d. Initiating recovery.\n",
290 			mddev->bitmap_info.cluster_name,
291 			slot->nodeid, slot->slot,
292 			cinfo->slot_number);
293 	set_bit(slot->slot - 1, &cinfo->recovery_map);
294 	if (!cinfo->recovery_thread) {
295 		cinfo->recovery_thread = md_register_thread(recover_bitmaps,
296 				mddev, "recover");
297 		if (!cinfo->recovery_thread) {
298 			pr_warn("md-cluster: Could not create recovery thread\n");
299 			return;
300 		}
301 	}
302 	md_wakeup_thread(cinfo->recovery_thread);
303 }
304 
305 static void recover_done(void *arg, struct dlm_slot *slots,
306 		int num_slots, int our_slot,
307 		uint32_t generation)
308 {
309 	struct mddev *mddev = arg;
310 	struct md_cluster_info *cinfo = mddev->cluster_info;
311 
312 	cinfo->slot_number = our_slot;
313 	complete(&cinfo->completion);
314 	clear_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state);
315 }
316 
317 static const struct dlm_lockspace_ops md_ls_ops = {
318 	.recover_prep = recover_prep,
319 	.recover_slot = recover_slot,
320 	.recover_done = recover_done,
321 };
322 
323 /*
324  * The BAST function for the ack lock resource
325  * This function wakes up the receive thread in
326  * order to receive and process the message.
327  */
328 static void ack_bast(void *arg, int mode)
329 {
330 	struct dlm_lock_resource *res = (struct dlm_lock_resource *)arg;
331 	struct md_cluster_info *cinfo = res->mddev->cluster_info;
332 
333 	if (mode == DLM_LOCK_EX)
334 		md_wakeup_thread(cinfo->recv_thread);
335 }
336 
337 static void __remove_suspend_info(struct md_cluster_info *cinfo, int slot)
338 {
339 	struct suspend_info *s, *tmp;
340 
341 	list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list)
342 		if (slot == s->slot) {
343 			pr_info("%s:%d Deleting suspend_info: %d\n",
344 					__func__, __LINE__, slot);
345 			list_del(&s->list);
346 			kfree(s);
347 			break;
348 		}
349 }
350 
351 static void remove_suspend_info(struct md_cluster_info *cinfo, int slot)
352 {
353 	spin_lock_irq(&cinfo->suspend_lock);
354 	__remove_suspend_info(cinfo, slot);
355 	spin_unlock_irq(&cinfo->suspend_lock);
356 }
357 
358 
359 static void process_suspend_info(struct md_cluster_info *cinfo,
360 		int slot, sector_t lo, sector_t hi)
361 {
362 	struct suspend_info *s;
363 
364 	if (!hi) {
365 		remove_suspend_info(cinfo, slot);
366 		return;
367 	}
368 	s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL);
369 	if (!s)
370 		return;
371 	s->slot = slot;
372 	s->lo = lo;
373 	s->hi = hi;
374 	spin_lock_irq(&cinfo->suspend_lock);
375 	/* Remove existing entry (if exists) before adding */
376 	__remove_suspend_info(cinfo, slot);
377 	list_add(&s->list, &cinfo->suspend_list);
378 	spin_unlock_irq(&cinfo->suspend_lock);
379 }
380 
381 static void process_add_new_disk(struct mddev *mddev, struct cluster_msg *cmsg)
382 {
383 	char disk_uuid[64];
384 	struct md_cluster_info *cinfo = mddev->cluster_info;
385 	char event_name[] = "EVENT=ADD_DEVICE";
386 	char raid_slot[16];
387 	char *envp[] = {event_name, disk_uuid, raid_slot, NULL};
388 	int len;
389 
390 	len = snprintf(disk_uuid, 64, "DEVICE_UUID=");
391 	pretty_uuid(disk_uuid + len, cmsg->uuid);
392 	snprintf(raid_slot, 16, "RAID_DISK=%d", cmsg->raid_slot);
393 	pr_info("%s:%d Sending kobject change with %s and %s\n", __func__, __LINE__, disk_uuid, raid_slot);
394 	init_completion(&cinfo->newdisk_completion);
395 	set_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
396 	kobject_uevent_env(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE, envp);
397 	wait_for_completion_timeout(&cinfo->newdisk_completion,
398 			NEW_DEV_TIMEOUT);
399 	clear_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
400 }
401 
402 
403 static void process_metadata_update(struct mddev *mddev, struct cluster_msg *msg)
404 {
405 	struct md_cluster_info *cinfo = mddev->cluster_info;
406 
407 	md_reload_sb(mddev);
408 	dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
409 }
410 
411 static void process_remove_disk(struct mddev *mddev, struct cluster_msg *msg)
412 {
413 	struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev, msg->raid_slot);
414 
415 	if (rdev)
416 		md_kick_rdev_from_array(rdev);
417 	else
418 		pr_warn("%s: %d Could not find disk(%d) to REMOVE\n", __func__, __LINE__, msg->raid_slot);
419 }
420 
421 static void process_readd_disk(struct mddev *mddev, struct cluster_msg *msg)
422 {
423 	struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev, msg->raid_slot);
424 
425 	if (rdev && test_bit(Faulty, &rdev->flags))
426 		clear_bit(Faulty, &rdev->flags);
427 	else
428 		pr_warn("%s: %d Could not find disk(%d) which is faulty", __func__, __LINE__, msg->raid_slot);
429 }
430 
431 static void process_recvd_msg(struct mddev *mddev, struct cluster_msg *msg)
432 {
433 	switch (msg->type) {
434 	case METADATA_UPDATED:
435 		pr_info("%s: %d Received message: METADATA_UPDATE from %d\n",
436 			__func__, __LINE__, msg->slot);
437 		process_metadata_update(mddev, msg);
438 		break;
439 	case RESYNCING:
440 		pr_info("%s: %d Received message: RESYNCING from %d\n",
441 			__func__, __LINE__, msg->slot);
442 		process_suspend_info(mddev->cluster_info, msg->slot,
443 				msg->low, msg->high);
444 		break;
445 	case NEWDISK:
446 		pr_info("%s: %d Received message: NEWDISK from %d\n",
447 			__func__, __LINE__, msg->slot);
448 		process_add_new_disk(mddev, msg);
449 		break;
450 	case REMOVE:
451 		pr_info("%s: %d Received REMOVE from %d\n",
452 			__func__, __LINE__, msg->slot);
453 		process_remove_disk(mddev, msg);
454 		break;
455 	case RE_ADD:
456 		pr_info("%s: %d Received RE_ADD from %d\n",
457 			__func__, __LINE__, msg->slot);
458 		process_readd_disk(mddev, msg);
459 		break;
460 	default:
461 		pr_warn("%s:%d Received unknown message from %d\n",
462 			__func__, __LINE__, msg->slot);
463 	}
464 }
465 
466 /*
467  * thread for receiving message
468  */
469 static void recv_daemon(struct md_thread *thread)
470 {
471 	struct md_cluster_info *cinfo = thread->mddev->cluster_info;
472 	struct dlm_lock_resource *ack_lockres = cinfo->ack_lockres;
473 	struct dlm_lock_resource *message_lockres = cinfo->message_lockres;
474 	struct cluster_msg msg;
475 
476 	/*get CR on Message*/
477 	if (dlm_lock_sync(message_lockres, DLM_LOCK_CR)) {
478 		pr_err("md/raid1:failed to get CR on MESSAGE\n");
479 		return;
480 	}
481 
482 	/* read lvb and wake up thread to process this message_lockres */
483 	memcpy(&msg, message_lockres->lksb.sb_lvbptr, sizeof(struct cluster_msg));
484 	process_recvd_msg(thread->mddev, &msg);
485 
486 	/*release CR on ack_lockres*/
487 	dlm_unlock_sync(ack_lockres);
488 	/*up-convert to EX on message_lockres*/
489 	dlm_lock_sync(message_lockres, DLM_LOCK_EX);
490 	/*get CR on ack_lockres again*/
491 	dlm_lock_sync(ack_lockres, DLM_LOCK_CR);
492 	/*release CR on message_lockres*/
493 	dlm_unlock_sync(message_lockres);
494 }
495 
496 /* lock_comm()
497  * Takes the lock on the TOKEN lock resource so no other
498  * node can communicate while the operation is underway.
499  */
500 static int lock_comm(struct md_cluster_info *cinfo)
501 {
502 	int error;
503 
504 	error = dlm_lock_sync(cinfo->token_lockres, DLM_LOCK_EX);
505 	if (error)
506 		pr_err("md-cluster(%s:%d): failed to get EX on TOKEN (%d)\n",
507 				__func__, __LINE__, error);
508 	return error;
509 }
510 
511 static void unlock_comm(struct md_cluster_info *cinfo)
512 {
513 	dlm_unlock_sync(cinfo->token_lockres);
514 }
515 
516 /* __sendmsg()
517  * This function performs the actual sending of the message. This function is
518  * usually called after performing the encompassing operation
519  * The function:
520  * 1. Grabs the message lockresource in EX mode
521  * 2. Copies the message to the message LVB
522  * 3. Downconverts message lockresource to CR
523  * 4. Upconverts ack lock resource from CR to EX. This forces the BAST on other nodes
524  *    and the other nodes read the message. The thread will wait here until all other
525  *    nodes have released ack lock resource.
526  * 5. Downconvert ack lockresource to CR
527  */
528 static int __sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
529 {
530 	int error;
531 	int slot = cinfo->slot_number - 1;
532 
533 	cmsg->slot = cpu_to_le32(slot);
534 	/*get EX on Message*/
535 	error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_EX);
536 	if (error) {
537 		pr_err("md-cluster: failed to get EX on MESSAGE (%d)\n", error);
538 		goto failed_message;
539 	}
540 
541 	memcpy(cinfo->message_lockres->lksb.sb_lvbptr, (void *)cmsg,
542 			sizeof(struct cluster_msg));
543 	/*down-convert EX to CR on Message*/
544 	error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_CR);
545 	if (error) {
546 		pr_err("md-cluster: failed to convert EX to CR on MESSAGE(%d)\n",
547 				error);
548 		goto failed_message;
549 	}
550 
551 	/*up-convert CR to EX on Ack*/
552 	error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_EX);
553 	if (error) {
554 		pr_err("md-cluster: failed to convert CR to EX on ACK(%d)\n",
555 				error);
556 		goto failed_ack;
557 	}
558 
559 	/*down-convert EX to CR on Ack*/
560 	error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR);
561 	if (error) {
562 		pr_err("md-cluster: failed to convert EX to CR on ACK(%d)\n",
563 				error);
564 		goto failed_ack;
565 	}
566 
567 failed_ack:
568 	dlm_unlock_sync(cinfo->message_lockres);
569 failed_message:
570 	return error;
571 }
572 
573 static int sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
574 {
575 	int ret;
576 
577 	lock_comm(cinfo);
578 	ret = __sendmsg(cinfo, cmsg);
579 	unlock_comm(cinfo);
580 	return ret;
581 }
582 
583 static int gather_all_resync_info(struct mddev *mddev, int total_slots)
584 {
585 	struct md_cluster_info *cinfo = mddev->cluster_info;
586 	int i, ret = 0;
587 	struct dlm_lock_resource *bm_lockres;
588 	struct suspend_info *s;
589 	char str[64];
590 
591 
592 	for (i = 0; i < total_slots; i++) {
593 		memset(str, '\0', 64);
594 		snprintf(str, 64, "bitmap%04d", i);
595 		bm_lockres = lockres_init(mddev, str, NULL, 1);
596 		if (!bm_lockres)
597 			return -ENOMEM;
598 		if (i == (cinfo->slot_number - 1))
599 			continue;
600 
601 		bm_lockres->flags |= DLM_LKF_NOQUEUE;
602 		ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
603 		if (ret == -EAGAIN) {
604 			memset(bm_lockres->lksb.sb_lvbptr, '\0', LVB_SIZE);
605 			s = read_resync_info(mddev, bm_lockres);
606 			if (s) {
607 				pr_info("%s:%d Resync[%llu..%llu] in progress on %d\n",
608 						__func__, __LINE__,
609 						(unsigned long long) s->lo,
610 						(unsigned long long) s->hi, i);
611 				spin_lock_irq(&cinfo->suspend_lock);
612 				s->slot = i;
613 				list_add(&s->list, &cinfo->suspend_list);
614 				spin_unlock_irq(&cinfo->suspend_lock);
615 			}
616 			ret = 0;
617 			lockres_free(bm_lockres);
618 			continue;
619 		}
620 		if (ret)
621 			goto out;
622 		/* TODO: Read the disk bitmap sb and check if it needs recovery */
623 		dlm_unlock_sync(bm_lockres);
624 		lockres_free(bm_lockres);
625 	}
626 out:
627 	return ret;
628 }
629 
630 static int join(struct mddev *mddev, int nodes)
631 {
632 	struct md_cluster_info *cinfo;
633 	int ret, ops_rv;
634 	char str[64];
635 
636 	if (!try_module_get(THIS_MODULE))
637 		return -ENOENT;
638 
639 	cinfo = kzalloc(sizeof(struct md_cluster_info), GFP_KERNEL);
640 	if (!cinfo)
641 		return -ENOMEM;
642 
643 	init_completion(&cinfo->completion);
644 
645 	mutex_init(&cinfo->sb_mutex);
646 	mddev->cluster_info = cinfo;
647 
648 	memset(str, 0, 64);
649 	pretty_uuid(str, mddev->uuid);
650 	ret = dlm_new_lockspace(str, mddev->bitmap_info.cluster_name,
651 				DLM_LSFL_FS, LVB_SIZE,
652 				&md_ls_ops, mddev, &ops_rv, &cinfo->lockspace);
653 	if (ret)
654 		goto err;
655 	wait_for_completion(&cinfo->completion);
656 	if (nodes < cinfo->slot_number) {
657 		pr_err("md-cluster: Slot allotted(%d) is greater than available slots(%d).",
658 			cinfo->slot_number, nodes);
659 		ret = -ERANGE;
660 		goto err;
661 	}
662 	cinfo->sb_lock = lockres_init(mddev, "cmd-super",
663 					NULL, 0);
664 	if (!cinfo->sb_lock) {
665 		ret = -ENOMEM;
666 		goto err;
667 	}
668 	/* Initiate the communication resources */
669 	ret = -ENOMEM;
670 	cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv");
671 	if (!cinfo->recv_thread) {
672 		pr_err("md-cluster: cannot allocate memory for recv_thread!\n");
673 		goto err;
674 	}
675 	cinfo->message_lockres = lockres_init(mddev, "message", NULL, 1);
676 	if (!cinfo->message_lockres)
677 		goto err;
678 	cinfo->token_lockres = lockres_init(mddev, "token", NULL, 0);
679 	if (!cinfo->token_lockres)
680 		goto err;
681 	cinfo->ack_lockres = lockres_init(mddev, "ack", ack_bast, 0);
682 	if (!cinfo->ack_lockres)
683 		goto err;
684 	cinfo->no_new_dev_lockres = lockres_init(mddev, "no-new-dev", NULL, 0);
685 	if (!cinfo->no_new_dev_lockres)
686 		goto err;
687 
688 	/* get sync CR lock on ACK. */
689 	if (dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR))
690 		pr_err("md-cluster: failed to get a sync CR lock on ACK!(%d)\n",
691 				ret);
692 	/* get sync CR lock on no-new-dev. */
693 	if (dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR))
694 		pr_err("md-cluster: failed to get a sync CR lock on no-new-dev!(%d)\n", ret);
695 
696 
697 	pr_info("md-cluster: Joined cluster %s slot %d\n", str, cinfo->slot_number);
698 	snprintf(str, 64, "bitmap%04d", cinfo->slot_number - 1);
699 	cinfo->bitmap_lockres = lockres_init(mddev, str, NULL, 1);
700 	if (!cinfo->bitmap_lockres)
701 		goto err;
702 	if (dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW)) {
703 		pr_err("Failed to get bitmap lock\n");
704 		ret = -EINVAL;
705 		goto err;
706 	}
707 
708 	INIT_LIST_HEAD(&cinfo->suspend_list);
709 	spin_lock_init(&cinfo->suspend_lock);
710 
711 	ret = gather_all_resync_info(mddev, nodes);
712 	if (ret)
713 		goto err;
714 
715 	return 0;
716 err:
717 	lockres_free(cinfo->message_lockres);
718 	lockres_free(cinfo->token_lockres);
719 	lockres_free(cinfo->ack_lockres);
720 	lockres_free(cinfo->no_new_dev_lockres);
721 	lockres_free(cinfo->bitmap_lockres);
722 	lockres_free(cinfo->sb_lock);
723 	if (cinfo->lockspace)
724 		dlm_release_lockspace(cinfo->lockspace, 2);
725 	mddev->cluster_info = NULL;
726 	kfree(cinfo);
727 	module_put(THIS_MODULE);
728 	return ret;
729 }
730 
731 static int leave(struct mddev *mddev)
732 {
733 	struct md_cluster_info *cinfo = mddev->cluster_info;
734 
735 	if (!cinfo)
736 		return 0;
737 	md_unregister_thread(&cinfo->recovery_thread);
738 	md_unregister_thread(&cinfo->recv_thread);
739 	lockres_free(cinfo->message_lockres);
740 	lockres_free(cinfo->token_lockres);
741 	lockres_free(cinfo->ack_lockres);
742 	lockres_free(cinfo->no_new_dev_lockres);
743 	lockres_free(cinfo->sb_lock);
744 	lockres_free(cinfo->bitmap_lockres);
745 	dlm_release_lockspace(cinfo->lockspace, 2);
746 	return 0;
747 }
748 
749 /* slot_number(): Returns the MD slot number to use
750  * DLM starts the slot numbers from 1, wheras cluster-md
751  * wants the number to be from zero, so we deduct one
752  */
753 static int slot_number(struct mddev *mddev)
754 {
755 	struct md_cluster_info *cinfo = mddev->cluster_info;
756 
757 	return cinfo->slot_number - 1;
758 }
759 
760 static void resync_info_update(struct mddev *mddev, sector_t lo, sector_t hi)
761 {
762 	struct md_cluster_info *cinfo = mddev->cluster_info;
763 
764 	add_resync_info(mddev, cinfo->bitmap_lockres, lo, hi);
765 	/* Re-acquire the lock to refresh LVB */
766 	dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW);
767 }
768 
769 static int metadata_update_start(struct mddev *mddev)
770 {
771 	return lock_comm(mddev->cluster_info);
772 }
773 
774 static int metadata_update_finish(struct mddev *mddev)
775 {
776 	struct md_cluster_info *cinfo = mddev->cluster_info;
777 	struct cluster_msg cmsg;
778 	int ret;
779 
780 	memset(&cmsg, 0, sizeof(cmsg));
781 	cmsg.type = cpu_to_le32(METADATA_UPDATED);
782 	ret = __sendmsg(cinfo, &cmsg);
783 	unlock_comm(cinfo);
784 	return ret;
785 }
786 
787 static int metadata_update_cancel(struct mddev *mddev)
788 {
789 	struct md_cluster_info *cinfo = mddev->cluster_info;
790 
791 	return dlm_unlock_sync(cinfo->token_lockres);
792 }
793 
794 static int resync_send(struct mddev *mddev, enum msg_type type,
795 		sector_t lo, sector_t hi)
796 {
797 	struct md_cluster_info *cinfo = mddev->cluster_info;
798 	struct cluster_msg cmsg;
799 	int slot = cinfo->slot_number - 1;
800 
801 	pr_info("%s:%d lo: %llu hi: %llu\n", __func__, __LINE__,
802 			(unsigned long long)lo,
803 			(unsigned long long)hi);
804 	resync_info_update(mddev, lo, hi);
805 	cmsg.type = cpu_to_le32(type);
806 	cmsg.slot = cpu_to_le32(slot);
807 	cmsg.low = cpu_to_le64(lo);
808 	cmsg.high = cpu_to_le64(hi);
809 	return sendmsg(cinfo, &cmsg);
810 }
811 
812 static int resync_start(struct mddev *mddev, sector_t lo, sector_t hi)
813 {
814 	pr_info("%s:%d\n", __func__, __LINE__);
815 	return resync_send(mddev, RESYNCING, lo, hi);
816 }
817 
818 static void resync_finish(struct mddev *mddev)
819 {
820 	pr_info("%s:%d\n", __func__, __LINE__);
821 	resync_send(mddev, RESYNCING, 0, 0);
822 }
823 
824 static int area_resyncing(struct mddev *mddev, int direction,
825 		sector_t lo, sector_t hi)
826 {
827 	struct md_cluster_info *cinfo = mddev->cluster_info;
828 	int ret = 0;
829 	struct suspend_info *s;
830 
831 	if ((direction == READ) &&
832 		test_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state))
833 		return 1;
834 
835 	spin_lock_irq(&cinfo->suspend_lock);
836 	if (list_empty(&cinfo->suspend_list))
837 		goto out;
838 	list_for_each_entry(s, &cinfo->suspend_list, list)
839 		if (hi > s->lo && lo < s->hi) {
840 			ret = 1;
841 			break;
842 		}
843 out:
844 	spin_unlock_irq(&cinfo->suspend_lock);
845 	return ret;
846 }
847 
848 static int add_new_disk_start(struct mddev *mddev, struct md_rdev *rdev)
849 {
850 	struct md_cluster_info *cinfo = mddev->cluster_info;
851 	struct cluster_msg cmsg;
852 	int ret = 0;
853 	struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
854 	char *uuid = sb->device_uuid;
855 
856 	memset(&cmsg, 0, sizeof(cmsg));
857 	cmsg.type = cpu_to_le32(NEWDISK);
858 	memcpy(cmsg.uuid, uuid, 16);
859 	cmsg.raid_slot = rdev->desc_nr;
860 	lock_comm(cinfo);
861 	ret = __sendmsg(cinfo, &cmsg);
862 	if (ret)
863 		return ret;
864 	cinfo->no_new_dev_lockres->flags |= DLM_LKF_NOQUEUE;
865 	ret = dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_EX);
866 	cinfo->no_new_dev_lockres->flags &= ~DLM_LKF_NOQUEUE;
867 	/* Some node does not "see" the device */
868 	if (ret == -EAGAIN)
869 		ret = -ENOENT;
870 	else
871 		dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
872 	return ret;
873 }
874 
875 static int add_new_disk_finish(struct mddev *mddev)
876 {
877 	struct cluster_msg cmsg;
878 	struct md_cluster_info *cinfo = mddev->cluster_info;
879 	int ret;
880 	/* Write sb and inform others */
881 	md_update_sb(mddev, 1);
882 	cmsg.type = METADATA_UPDATED;
883 	ret = __sendmsg(cinfo, &cmsg);
884 	unlock_comm(cinfo);
885 	return ret;
886 }
887 
888 static int new_disk_ack(struct mddev *mddev, bool ack)
889 {
890 	struct md_cluster_info *cinfo = mddev->cluster_info;
891 
892 	if (!test_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state)) {
893 		pr_warn("md-cluster(%s): Spurious cluster confirmation\n", mdname(mddev));
894 		return -EINVAL;
895 	}
896 
897 	if (ack)
898 		dlm_unlock_sync(cinfo->no_new_dev_lockres);
899 	complete(&cinfo->newdisk_completion);
900 	return 0;
901 }
902 
903 static int remove_disk(struct mddev *mddev, struct md_rdev *rdev)
904 {
905 	struct cluster_msg cmsg;
906 	struct md_cluster_info *cinfo = mddev->cluster_info;
907 	cmsg.type = REMOVE;
908 	cmsg.raid_slot = rdev->desc_nr;
909 	return __sendmsg(cinfo, &cmsg);
910 }
911 
912 static int gather_bitmaps(struct md_rdev *rdev)
913 {
914 	int sn, err;
915 	sector_t lo, hi;
916 	struct cluster_msg cmsg;
917 	struct mddev *mddev = rdev->mddev;
918 	struct md_cluster_info *cinfo = mddev->cluster_info;
919 
920 	cmsg.type = RE_ADD;
921 	cmsg.raid_slot = rdev->desc_nr;
922 	err = sendmsg(cinfo, &cmsg);
923 	if (err)
924 		goto out;
925 
926 	for (sn = 0; sn < mddev->bitmap_info.nodes; sn++) {
927 		if (sn == (cinfo->slot_number - 1))
928 			continue;
929 		err = bitmap_copy_from_slot(mddev, sn, &lo, &hi, false);
930 		if (err) {
931 			pr_warn("md-cluster: Could not gather bitmaps from slot %d", sn);
932 			goto out;
933 		}
934 		if ((hi > 0) && (lo < mddev->recovery_cp))
935 			mddev->recovery_cp = lo;
936 	}
937 out:
938 	return err;
939 }
940 
941 static struct md_cluster_operations cluster_ops = {
942 	.join   = join,
943 	.leave  = leave,
944 	.slot_number = slot_number,
945 	.resync_info_update = resync_info_update,
946 	.resync_start = resync_start,
947 	.resync_finish = resync_finish,
948 	.metadata_update_start = metadata_update_start,
949 	.metadata_update_finish = metadata_update_finish,
950 	.metadata_update_cancel = metadata_update_cancel,
951 	.area_resyncing = area_resyncing,
952 	.add_new_disk_start = add_new_disk_start,
953 	.add_new_disk_finish = add_new_disk_finish,
954 	.new_disk_ack = new_disk_ack,
955 	.remove_disk = remove_disk,
956 	.gather_bitmaps = gather_bitmaps,
957 };
958 
959 static int __init cluster_init(void)
960 {
961 	pr_warn("md-cluster: EXPERIMENTAL. Use with caution\n");
962 	pr_info("Registering Cluster MD functions\n");
963 	register_md_cluster_operations(&cluster_ops, THIS_MODULE);
964 	return 0;
965 }
966 
967 static void cluster_exit(void)
968 {
969 	unregister_md_cluster_operations();
970 }
971 
972 module_init(cluster_init);
973 module_exit(cluster_exit);
974 MODULE_LICENSE("GPL");
975 MODULE_DESCRIPTION("Clustering support for MD");
976