xref: /openbmc/linux/drivers/md/md-cluster.c (revision 4bce6fce)
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 
48 
49 struct md_cluster_info {
50 	/* dlm lock space and resources for clustered raid. */
51 	dlm_lockspace_t *lockspace;
52 	int slot_number;
53 	struct completion completion;
54 	struct dlm_lock_resource *sb_lock;
55 	struct mutex sb_mutex;
56 	struct dlm_lock_resource *bitmap_lockres;
57 	struct list_head suspend_list;
58 	spinlock_t suspend_lock;
59 	struct md_thread *recovery_thread;
60 	unsigned long recovery_map;
61 	/* communication loc resources */
62 	struct dlm_lock_resource *ack_lockres;
63 	struct dlm_lock_resource *message_lockres;
64 	struct dlm_lock_resource *token_lockres;
65 	struct dlm_lock_resource *no_new_dev_lockres;
66 	struct md_thread *recv_thread;
67 	struct completion newdisk_completion;
68 	unsigned long state;
69 };
70 
71 enum msg_type {
72 	METADATA_UPDATED = 0,
73 	RESYNCING,
74 	NEWDISK,
75 	REMOVE,
76 	RE_ADD,
77 };
78 
79 struct cluster_msg {
80 	int type;
81 	int slot;
82 	/* TODO: Unionize this for smaller footprint */
83 	sector_t low;
84 	sector_t high;
85 	char uuid[16];
86 	int raid_slot;
87 };
88 
89 static void sync_ast(void *arg)
90 {
91 	struct dlm_lock_resource *res;
92 
93 	res = (struct dlm_lock_resource *) arg;
94 	complete(&res->completion);
95 }
96 
97 static int dlm_lock_sync(struct dlm_lock_resource *res, int mode)
98 {
99 	int ret = 0;
100 
101 	init_completion(&res->completion);
102 	ret = dlm_lock(res->ls, mode, &res->lksb,
103 			res->flags, res->name, strlen(res->name),
104 			0, sync_ast, res, res->bast);
105 	if (ret)
106 		return ret;
107 	wait_for_completion(&res->completion);
108 	return res->lksb.sb_status;
109 }
110 
111 static int dlm_unlock_sync(struct dlm_lock_resource *res)
112 {
113 	return dlm_lock_sync(res, DLM_LOCK_NL);
114 }
115 
116 static struct dlm_lock_resource *lockres_init(struct mddev *mddev,
117 		char *name, void (*bastfn)(void *arg, int mode), int with_lvb)
118 {
119 	struct dlm_lock_resource *res = NULL;
120 	int ret, namelen;
121 	struct md_cluster_info *cinfo = mddev->cluster_info;
122 
123 	res = kzalloc(sizeof(struct dlm_lock_resource), GFP_KERNEL);
124 	if (!res)
125 		return NULL;
126 	res->ls = cinfo->lockspace;
127 	res->mddev = mddev;
128 	namelen = strlen(name);
129 	res->name = kzalloc(namelen + 1, GFP_KERNEL);
130 	if (!res->name) {
131 		pr_err("md-cluster: Unable to allocate resource name for resource %s\n", name);
132 		goto out_err;
133 	}
134 	strlcpy(res->name, name, namelen + 1);
135 	if (with_lvb) {
136 		res->lksb.sb_lvbptr = kzalloc(LVB_SIZE, GFP_KERNEL);
137 		if (!res->lksb.sb_lvbptr) {
138 			pr_err("md-cluster: Unable to allocate LVB for resource %s\n", name);
139 			goto out_err;
140 		}
141 		res->flags = DLM_LKF_VALBLK;
142 	}
143 
144 	if (bastfn)
145 		res->bast = bastfn;
146 
147 	res->flags |= DLM_LKF_EXPEDITE;
148 
149 	ret = dlm_lock_sync(res, DLM_LOCK_NL);
150 	if (ret) {
151 		pr_err("md-cluster: Unable to lock NL on new lock resource %s\n", name);
152 		goto out_err;
153 	}
154 	res->flags &= ~DLM_LKF_EXPEDITE;
155 	res->flags |= DLM_LKF_CONVERT;
156 
157 	return res;
158 out_err:
159 	kfree(res->lksb.sb_lvbptr);
160 	kfree(res->name);
161 	kfree(res);
162 	return NULL;
163 }
164 
165 static void lockres_free(struct dlm_lock_resource *res)
166 {
167 	if (!res)
168 		return;
169 
170 	init_completion(&res->completion);
171 	dlm_unlock(res->ls, res->lksb.sb_lkid, 0, &res->lksb, res);
172 	wait_for_completion(&res->completion);
173 
174 	kfree(res->name);
175 	kfree(res->lksb.sb_lvbptr);
176 	kfree(res);
177 }
178 
179 static char *pretty_uuid(char *dest, char *src)
180 {
181 	int i, len = 0;
182 
183 	for (i = 0; i < 16; i++) {
184 		if (i == 4 || i == 6 || i == 8 || i == 10)
185 			len += sprintf(dest + len, "-");
186 		len += sprintf(dest + len, "%02x", (__u8)src[i]);
187 	}
188 	return dest;
189 }
190 
191 static void add_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres,
192 		sector_t lo, sector_t hi)
193 {
194 	struct resync_info *ri;
195 
196 	ri = (struct resync_info *)lockres->lksb.sb_lvbptr;
197 	ri->lo = cpu_to_le64(lo);
198 	ri->hi = cpu_to_le64(hi);
199 }
200 
201 static struct suspend_info *read_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres)
202 {
203 	struct resync_info ri;
204 	struct suspend_info *s = NULL;
205 	sector_t hi = 0;
206 
207 	dlm_lock_sync(lockres, DLM_LOCK_CR);
208 	memcpy(&ri, lockres->lksb.sb_lvbptr, sizeof(struct resync_info));
209 	hi = le64_to_cpu(ri.hi);
210 	if (ri.hi > 0) {
211 		s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL);
212 		if (!s)
213 			goto out;
214 		s->hi = hi;
215 		s->lo = le64_to_cpu(ri.lo);
216 	}
217 	dlm_unlock_sync(lockres);
218 out:
219 	return s;
220 }
221 
222 static void recover_bitmaps(struct md_thread *thread)
223 {
224 	struct mddev *mddev = thread->mddev;
225 	struct md_cluster_info *cinfo = mddev->cluster_info;
226 	struct dlm_lock_resource *bm_lockres;
227 	char str[64];
228 	int slot, ret;
229 	struct suspend_info *s, *tmp;
230 	sector_t lo, hi;
231 
232 	while (cinfo->recovery_map) {
233 		slot = fls64((u64)cinfo->recovery_map) - 1;
234 
235 		/* Clear suspend_area associated with the bitmap */
236 		spin_lock_irq(&cinfo->suspend_lock);
237 		list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list)
238 			if (slot == s->slot) {
239 				list_del(&s->list);
240 				kfree(s);
241 			}
242 		spin_unlock_irq(&cinfo->suspend_lock);
243 
244 		snprintf(str, 64, "bitmap%04d", slot);
245 		bm_lockres = lockres_init(mddev, str, NULL, 1);
246 		if (!bm_lockres) {
247 			pr_err("md-cluster: Cannot initialize bitmaps\n");
248 			goto clear_bit;
249 		}
250 
251 		ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
252 		if (ret) {
253 			pr_err("md-cluster: Could not DLM lock %s: %d\n",
254 					str, ret);
255 			goto clear_bit;
256 		}
257 		ret = bitmap_copy_from_slot(mddev, slot, &lo, &hi, true);
258 		if (ret) {
259 			pr_err("md-cluster: Could not copy data from bitmap %d\n", slot);
260 			goto dlm_unlock;
261 		}
262 		if (hi > 0) {
263 			/* TODO:Wait for current resync to get over */
264 			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
265 			if (lo < mddev->recovery_cp)
266 				mddev->recovery_cp = lo;
267 			md_check_recovery(mddev);
268 		}
269 dlm_unlock:
270 		dlm_unlock_sync(bm_lockres);
271 clear_bit:
272 		clear_bit(slot, &cinfo->recovery_map);
273 	}
274 }
275 
276 static void recover_prep(void *arg)
277 {
278 }
279 
280 static void recover_slot(void *arg, struct dlm_slot *slot)
281 {
282 	struct mddev *mddev = arg;
283 	struct md_cluster_info *cinfo = mddev->cluster_info;
284 
285 	pr_info("md-cluster: %s Node %d/%d down. My slot: %d. Initiating recovery.\n",
286 			mddev->bitmap_info.cluster_name,
287 			slot->nodeid, slot->slot,
288 			cinfo->slot_number);
289 	set_bit(slot->slot - 1, &cinfo->recovery_map);
290 	if (!cinfo->recovery_thread) {
291 		cinfo->recovery_thread = md_register_thread(recover_bitmaps,
292 				mddev, "recover");
293 		if (!cinfo->recovery_thread) {
294 			pr_warn("md-cluster: Could not create recovery thread\n");
295 			return;
296 		}
297 	}
298 	md_wakeup_thread(cinfo->recovery_thread);
299 }
300 
301 static void recover_done(void *arg, struct dlm_slot *slots,
302 		int num_slots, int our_slot,
303 		uint32_t generation)
304 {
305 	struct mddev *mddev = arg;
306 	struct md_cluster_info *cinfo = mddev->cluster_info;
307 
308 	cinfo->slot_number = our_slot;
309 	complete(&cinfo->completion);
310 }
311 
312 static const struct dlm_lockspace_ops md_ls_ops = {
313 	.recover_prep = recover_prep,
314 	.recover_slot = recover_slot,
315 	.recover_done = recover_done,
316 };
317 
318 /*
319  * The BAST function for the ack lock resource
320  * This function wakes up the receive thread in
321  * order to receive and process the message.
322  */
323 static void ack_bast(void *arg, int mode)
324 {
325 	struct dlm_lock_resource *res = (struct dlm_lock_resource *)arg;
326 	struct md_cluster_info *cinfo = res->mddev->cluster_info;
327 
328 	if (mode == DLM_LOCK_EX)
329 		md_wakeup_thread(cinfo->recv_thread);
330 }
331 
332 static void __remove_suspend_info(struct md_cluster_info *cinfo, int slot)
333 {
334 	struct suspend_info *s, *tmp;
335 
336 	list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list)
337 		if (slot == s->slot) {
338 			pr_info("%s:%d Deleting suspend_info: %d\n",
339 					__func__, __LINE__, slot);
340 			list_del(&s->list);
341 			kfree(s);
342 			break;
343 		}
344 }
345 
346 static void remove_suspend_info(struct md_cluster_info *cinfo, int slot)
347 {
348 	spin_lock_irq(&cinfo->suspend_lock);
349 	__remove_suspend_info(cinfo, slot);
350 	spin_unlock_irq(&cinfo->suspend_lock);
351 }
352 
353 
354 static void process_suspend_info(struct md_cluster_info *cinfo,
355 		int slot, sector_t lo, sector_t hi)
356 {
357 	struct suspend_info *s;
358 
359 	if (!hi) {
360 		remove_suspend_info(cinfo, slot);
361 		return;
362 	}
363 	s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL);
364 	if (!s)
365 		return;
366 	s->slot = slot;
367 	s->lo = lo;
368 	s->hi = hi;
369 	spin_lock_irq(&cinfo->suspend_lock);
370 	/* Remove existing entry (if exists) before adding */
371 	__remove_suspend_info(cinfo, slot);
372 	list_add(&s->list, &cinfo->suspend_list);
373 	spin_unlock_irq(&cinfo->suspend_lock);
374 }
375 
376 static void process_add_new_disk(struct mddev *mddev, struct cluster_msg *cmsg)
377 {
378 	char disk_uuid[64];
379 	struct md_cluster_info *cinfo = mddev->cluster_info;
380 	char event_name[] = "EVENT=ADD_DEVICE";
381 	char raid_slot[16];
382 	char *envp[] = {event_name, disk_uuid, raid_slot, NULL};
383 	int len;
384 
385 	len = snprintf(disk_uuid, 64, "DEVICE_UUID=");
386 	pretty_uuid(disk_uuid + len, cmsg->uuid);
387 	snprintf(raid_slot, 16, "RAID_DISK=%d", cmsg->raid_slot);
388 	pr_info("%s:%d Sending kobject change with %s and %s\n", __func__, __LINE__, disk_uuid, raid_slot);
389 	init_completion(&cinfo->newdisk_completion);
390 	set_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
391 	kobject_uevent_env(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE, envp);
392 	wait_for_completion_timeout(&cinfo->newdisk_completion,
393 			NEW_DEV_TIMEOUT);
394 	clear_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
395 }
396 
397 
398 static void process_metadata_update(struct mddev *mddev, struct cluster_msg *msg)
399 {
400 	struct md_cluster_info *cinfo = mddev->cluster_info;
401 
402 	md_reload_sb(mddev);
403 	dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
404 }
405 
406 static void process_remove_disk(struct mddev *mddev, struct cluster_msg *msg)
407 {
408 	struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev, msg->raid_slot);
409 
410 	if (rdev)
411 		md_kick_rdev_from_array(rdev);
412 	else
413 		pr_warn("%s: %d Could not find disk(%d) to REMOVE\n", __func__, __LINE__, msg->raid_slot);
414 }
415 
416 static void process_readd_disk(struct mddev *mddev, struct cluster_msg *msg)
417 {
418 	struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev, msg->raid_slot);
419 
420 	if (rdev && test_bit(Faulty, &rdev->flags))
421 		clear_bit(Faulty, &rdev->flags);
422 	else
423 		pr_warn("%s: %d Could not find disk(%d) which is faulty", __func__, __LINE__, msg->raid_slot);
424 }
425 
426 static void process_recvd_msg(struct mddev *mddev, struct cluster_msg *msg)
427 {
428 	switch (msg->type) {
429 	case METADATA_UPDATED:
430 		pr_info("%s: %d Received message: METADATA_UPDATE from %d\n",
431 			__func__, __LINE__, msg->slot);
432 		process_metadata_update(mddev, msg);
433 		break;
434 	case RESYNCING:
435 		pr_info("%s: %d Received message: RESYNCING from %d\n",
436 			__func__, __LINE__, msg->slot);
437 		process_suspend_info(mddev->cluster_info, msg->slot,
438 				msg->low, msg->high);
439 		break;
440 	case NEWDISK:
441 		pr_info("%s: %d Received message: NEWDISK from %d\n",
442 			__func__, __LINE__, msg->slot);
443 		process_add_new_disk(mddev, msg);
444 		break;
445 	case REMOVE:
446 		pr_info("%s: %d Received REMOVE from %d\n",
447 			__func__, __LINE__, msg->slot);
448 		process_remove_disk(mddev, msg);
449 		break;
450 	case RE_ADD:
451 		pr_info("%s: %d Received RE_ADD from %d\n",
452 			__func__, __LINE__, msg->slot);
453 		process_readd_disk(mddev, msg);
454 		break;
455 	default:
456 		pr_warn("%s:%d Received unknown message from %d\n",
457 			__func__, __LINE__, msg->slot);
458 	}
459 }
460 
461 /*
462  * thread for receiving message
463  */
464 static void recv_daemon(struct md_thread *thread)
465 {
466 	struct md_cluster_info *cinfo = thread->mddev->cluster_info;
467 	struct dlm_lock_resource *ack_lockres = cinfo->ack_lockres;
468 	struct dlm_lock_resource *message_lockres = cinfo->message_lockres;
469 	struct cluster_msg msg;
470 
471 	/*get CR on Message*/
472 	if (dlm_lock_sync(message_lockres, DLM_LOCK_CR)) {
473 		pr_err("md/raid1:failed to get CR on MESSAGE\n");
474 		return;
475 	}
476 
477 	/* read lvb and wake up thread to process this message_lockres */
478 	memcpy(&msg, message_lockres->lksb.sb_lvbptr, sizeof(struct cluster_msg));
479 	process_recvd_msg(thread->mddev, &msg);
480 
481 	/*release CR on ack_lockres*/
482 	dlm_unlock_sync(ack_lockres);
483 	/*up-convert to EX on message_lockres*/
484 	dlm_lock_sync(message_lockres, DLM_LOCK_EX);
485 	/*get CR on ack_lockres again*/
486 	dlm_lock_sync(ack_lockres, DLM_LOCK_CR);
487 	/*release CR on message_lockres*/
488 	dlm_unlock_sync(message_lockres);
489 }
490 
491 /* lock_comm()
492  * Takes the lock on the TOKEN lock resource so no other
493  * node can communicate while the operation is underway.
494  */
495 static int lock_comm(struct md_cluster_info *cinfo)
496 {
497 	int error;
498 
499 	error = dlm_lock_sync(cinfo->token_lockres, DLM_LOCK_EX);
500 	if (error)
501 		pr_err("md-cluster(%s:%d): failed to get EX on TOKEN (%d)\n",
502 				__func__, __LINE__, error);
503 	return error;
504 }
505 
506 static void unlock_comm(struct md_cluster_info *cinfo)
507 {
508 	dlm_unlock_sync(cinfo->token_lockres);
509 }
510 
511 /* __sendmsg()
512  * This function performs the actual sending of the message. This function is
513  * usually called after performing the encompassing operation
514  * The function:
515  * 1. Grabs the message lockresource in EX mode
516  * 2. Copies the message to the message LVB
517  * 3. Downconverts message lockresource to CR
518  * 4. Upconverts ack lock resource from CR to EX. This forces the BAST on other nodes
519  *    and the other nodes read the message. The thread will wait here until all other
520  *    nodes have released ack lock resource.
521  * 5. Downconvert ack lockresource to CR
522  */
523 static int __sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
524 {
525 	int error;
526 	int slot = cinfo->slot_number - 1;
527 
528 	cmsg->slot = cpu_to_le32(slot);
529 	/*get EX on Message*/
530 	error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_EX);
531 	if (error) {
532 		pr_err("md-cluster: failed to get EX on MESSAGE (%d)\n", error);
533 		goto failed_message;
534 	}
535 
536 	memcpy(cinfo->message_lockres->lksb.sb_lvbptr, (void *)cmsg,
537 			sizeof(struct cluster_msg));
538 	/*down-convert EX to CR on Message*/
539 	error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_CR);
540 	if (error) {
541 		pr_err("md-cluster: failed to convert EX to CR on MESSAGE(%d)\n",
542 				error);
543 		goto failed_message;
544 	}
545 
546 	/*up-convert CR to EX on Ack*/
547 	error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_EX);
548 	if (error) {
549 		pr_err("md-cluster: failed to convert CR to EX on ACK(%d)\n",
550 				error);
551 		goto failed_ack;
552 	}
553 
554 	/*down-convert EX to CR on Ack*/
555 	error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR);
556 	if (error) {
557 		pr_err("md-cluster: failed to convert EX to CR on ACK(%d)\n",
558 				error);
559 		goto failed_ack;
560 	}
561 
562 failed_ack:
563 	dlm_unlock_sync(cinfo->message_lockres);
564 failed_message:
565 	return error;
566 }
567 
568 static int sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
569 {
570 	int ret;
571 
572 	lock_comm(cinfo);
573 	ret = __sendmsg(cinfo, cmsg);
574 	unlock_comm(cinfo);
575 	return ret;
576 }
577 
578 static int gather_all_resync_info(struct mddev *mddev, int total_slots)
579 {
580 	struct md_cluster_info *cinfo = mddev->cluster_info;
581 	int i, ret = 0;
582 	struct dlm_lock_resource *bm_lockres;
583 	struct suspend_info *s;
584 	char str[64];
585 
586 
587 	for (i = 0; i < total_slots; i++) {
588 		memset(str, '\0', 64);
589 		snprintf(str, 64, "bitmap%04d", i);
590 		bm_lockres = lockres_init(mddev, str, NULL, 1);
591 		if (!bm_lockres)
592 			return -ENOMEM;
593 		if (i == (cinfo->slot_number - 1))
594 			continue;
595 
596 		bm_lockres->flags |= DLM_LKF_NOQUEUE;
597 		ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
598 		if (ret == -EAGAIN) {
599 			memset(bm_lockres->lksb.sb_lvbptr, '\0', LVB_SIZE);
600 			s = read_resync_info(mddev, bm_lockres);
601 			if (s) {
602 				pr_info("%s:%d Resync[%llu..%llu] in progress on %d\n",
603 						__func__, __LINE__,
604 						(unsigned long long) s->lo,
605 						(unsigned long long) s->hi, i);
606 				spin_lock_irq(&cinfo->suspend_lock);
607 				s->slot = i;
608 				list_add(&s->list, &cinfo->suspend_list);
609 				spin_unlock_irq(&cinfo->suspend_lock);
610 			}
611 			ret = 0;
612 			lockres_free(bm_lockres);
613 			continue;
614 		}
615 		if (ret)
616 			goto out;
617 		/* TODO: Read the disk bitmap sb and check if it needs recovery */
618 		dlm_unlock_sync(bm_lockres);
619 		lockres_free(bm_lockres);
620 	}
621 out:
622 	return ret;
623 }
624 
625 static int join(struct mddev *mddev, int nodes)
626 {
627 	struct md_cluster_info *cinfo;
628 	int ret, ops_rv;
629 	char str[64];
630 
631 	if (!try_module_get(THIS_MODULE))
632 		return -ENOENT;
633 
634 	cinfo = kzalloc(sizeof(struct md_cluster_info), GFP_KERNEL);
635 	if (!cinfo)
636 		return -ENOMEM;
637 
638 	init_completion(&cinfo->completion);
639 
640 	mutex_init(&cinfo->sb_mutex);
641 	mddev->cluster_info = cinfo;
642 
643 	memset(str, 0, 64);
644 	pretty_uuid(str, mddev->uuid);
645 	ret = dlm_new_lockspace(str, mddev->bitmap_info.cluster_name,
646 				DLM_LSFL_FS, LVB_SIZE,
647 				&md_ls_ops, mddev, &ops_rv, &cinfo->lockspace);
648 	if (ret)
649 		goto err;
650 	wait_for_completion(&cinfo->completion);
651 	if (nodes < cinfo->slot_number) {
652 		pr_err("md-cluster: Slot allotted(%d) is greater than available slots(%d).",
653 			cinfo->slot_number, nodes);
654 		ret = -ERANGE;
655 		goto err;
656 	}
657 	cinfo->sb_lock = lockres_init(mddev, "cmd-super",
658 					NULL, 0);
659 	if (!cinfo->sb_lock) {
660 		ret = -ENOMEM;
661 		goto err;
662 	}
663 	/* Initiate the communication resources */
664 	ret = -ENOMEM;
665 	cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv");
666 	if (!cinfo->recv_thread) {
667 		pr_err("md-cluster: cannot allocate memory for recv_thread!\n");
668 		goto err;
669 	}
670 	cinfo->message_lockres = lockres_init(mddev, "message", NULL, 1);
671 	if (!cinfo->message_lockres)
672 		goto err;
673 	cinfo->token_lockres = lockres_init(mddev, "token", NULL, 0);
674 	if (!cinfo->token_lockres)
675 		goto err;
676 	cinfo->ack_lockres = lockres_init(mddev, "ack", ack_bast, 0);
677 	if (!cinfo->ack_lockres)
678 		goto err;
679 	cinfo->no_new_dev_lockres = lockres_init(mddev, "no-new-dev", NULL, 0);
680 	if (!cinfo->no_new_dev_lockres)
681 		goto err;
682 
683 	/* get sync CR lock on ACK. */
684 	if (dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR))
685 		pr_err("md-cluster: failed to get a sync CR lock on ACK!(%d)\n",
686 				ret);
687 	/* get sync CR lock on no-new-dev. */
688 	if (dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR))
689 		pr_err("md-cluster: failed to get a sync CR lock on no-new-dev!(%d)\n", ret);
690 
691 
692 	pr_info("md-cluster: Joined cluster %s slot %d\n", str, cinfo->slot_number);
693 	snprintf(str, 64, "bitmap%04d", cinfo->slot_number - 1);
694 	cinfo->bitmap_lockres = lockres_init(mddev, str, NULL, 1);
695 	if (!cinfo->bitmap_lockres)
696 		goto err;
697 	if (dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW)) {
698 		pr_err("Failed to get bitmap lock\n");
699 		ret = -EINVAL;
700 		goto err;
701 	}
702 
703 	INIT_LIST_HEAD(&cinfo->suspend_list);
704 	spin_lock_init(&cinfo->suspend_lock);
705 
706 	ret = gather_all_resync_info(mddev, nodes);
707 	if (ret)
708 		goto err;
709 
710 	return 0;
711 err:
712 	lockres_free(cinfo->message_lockres);
713 	lockres_free(cinfo->token_lockres);
714 	lockres_free(cinfo->ack_lockres);
715 	lockres_free(cinfo->no_new_dev_lockres);
716 	lockres_free(cinfo->bitmap_lockres);
717 	lockres_free(cinfo->sb_lock);
718 	if (cinfo->lockspace)
719 		dlm_release_lockspace(cinfo->lockspace, 2);
720 	mddev->cluster_info = NULL;
721 	kfree(cinfo);
722 	module_put(THIS_MODULE);
723 	return ret;
724 }
725 
726 static int leave(struct mddev *mddev)
727 {
728 	struct md_cluster_info *cinfo = mddev->cluster_info;
729 
730 	if (!cinfo)
731 		return 0;
732 	md_unregister_thread(&cinfo->recovery_thread);
733 	md_unregister_thread(&cinfo->recv_thread);
734 	lockres_free(cinfo->message_lockres);
735 	lockres_free(cinfo->token_lockres);
736 	lockres_free(cinfo->ack_lockres);
737 	lockres_free(cinfo->no_new_dev_lockres);
738 	lockres_free(cinfo->sb_lock);
739 	lockres_free(cinfo->bitmap_lockres);
740 	dlm_release_lockspace(cinfo->lockspace, 2);
741 	return 0;
742 }
743 
744 /* slot_number(): Returns the MD slot number to use
745  * DLM starts the slot numbers from 1, wheras cluster-md
746  * wants the number to be from zero, so we deduct one
747  */
748 static int slot_number(struct mddev *mddev)
749 {
750 	struct md_cluster_info *cinfo = mddev->cluster_info;
751 
752 	return cinfo->slot_number - 1;
753 }
754 
755 static void resync_info_update(struct mddev *mddev, sector_t lo, sector_t hi)
756 {
757 	struct md_cluster_info *cinfo = mddev->cluster_info;
758 
759 	add_resync_info(mddev, cinfo->bitmap_lockres, lo, hi);
760 	/* Re-acquire the lock to refresh LVB */
761 	dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW);
762 }
763 
764 static int metadata_update_start(struct mddev *mddev)
765 {
766 	return lock_comm(mddev->cluster_info);
767 }
768 
769 static int metadata_update_finish(struct mddev *mddev)
770 {
771 	struct md_cluster_info *cinfo = mddev->cluster_info;
772 	struct cluster_msg cmsg;
773 	int ret;
774 
775 	memset(&cmsg, 0, sizeof(cmsg));
776 	cmsg.type = cpu_to_le32(METADATA_UPDATED);
777 	ret = __sendmsg(cinfo, &cmsg);
778 	unlock_comm(cinfo);
779 	return ret;
780 }
781 
782 static int metadata_update_cancel(struct mddev *mddev)
783 {
784 	struct md_cluster_info *cinfo = mddev->cluster_info;
785 
786 	return dlm_unlock_sync(cinfo->token_lockres);
787 }
788 
789 static int resync_send(struct mddev *mddev, enum msg_type type,
790 		sector_t lo, sector_t hi)
791 {
792 	struct md_cluster_info *cinfo = mddev->cluster_info;
793 	struct cluster_msg cmsg;
794 	int slot = cinfo->slot_number - 1;
795 
796 	pr_info("%s:%d lo: %llu hi: %llu\n", __func__, __LINE__,
797 			(unsigned long long)lo,
798 			(unsigned long long)hi);
799 	resync_info_update(mddev, lo, hi);
800 	cmsg.type = cpu_to_le32(type);
801 	cmsg.slot = cpu_to_le32(slot);
802 	cmsg.low = cpu_to_le64(lo);
803 	cmsg.high = cpu_to_le64(hi);
804 	return sendmsg(cinfo, &cmsg);
805 }
806 
807 static int resync_start(struct mddev *mddev, sector_t lo, sector_t hi)
808 {
809 	pr_info("%s:%d\n", __func__, __LINE__);
810 	return resync_send(mddev, RESYNCING, lo, hi);
811 }
812 
813 static void resync_finish(struct mddev *mddev)
814 {
815 	pr_info("%s:%d\n", __func__, __LINE__);
816 	resync_send(mddev, RESYNCING, 0, 0);
817 }
818 
819 static int area_resyncing(struct mddev *mddev, sector_t lo, sector_t hi)
820 {
821 	struct md_cluster_info *cinfo = mddev->cluster_info;
822 	int ret = 0;
823 	struct suspend_info *s;
824 
825 	spin_lock_irq(&cinfo->suspend_lock);
826 	if (list_empty(&cinfo->suspend_list))
827 		goto out;
828 	list_for_each_entry(s, &cinfo->suspend_list, list)
829 		if (hi > s->lo && lo < s->hi) {
830 			ret = 1;
831 			break;
832 		}
833 out:
834 	spin_unlock_irq(&cinfo->suspend_lock);
835 	return ret;
836 }
837 
838 static int add_new_disk_start(struct mddev *mddev, struct md_rdev *rdev)
839 {
840 	struct md_cluster_info *cinfo = mddev->cluster_info;
841 	struct cluster_msg cmsg;
842 	int ret = 0;
843 	struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
844 	char *uuid = sb->device_uuid;
845 
846 	memset(&cmsg, 0, sizeof(cmsg));
847 	cmsg.type = cpu_to_le32(NEWDISK);
848 	memcpy(cmsg.uuid, uuid, 16);
849 	cmsg.raid_slot = rdev->desc_nr;
850 	lock_comm(cinfo);
851 	ret = __sendmsg(cinfo, &cmsg);
852 	if (ret)
853 		return ret;
854 	cinfo->no_new_dev_lockres->flags |= DLM_LKF_NOQUEUE;
855 	ret = dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_EX);
856 	cinfo->no_new_dev_lockres->flags &= ~DLM_LKF_NOQUEUE;
857 	/* Some node does not "see" the device */
858 	if (ret == -EAGAIN)
859 		ret = -ENOENT;
860 	else
861 		dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
862 	return ret;
863 }
864 
865 static int add_new_disk_finish(struct mddev *mddev)
866 {
867 	struct cluster_msg cmsg;
868 	struct md_cluster_info *cinfo = mddev->cluster_info;
869 	int ret;
870 	/* Write sb and inform others */
871 	md_update_sb(mddev, 1);
872 	cmsg.type = METADATA_UPDATED;
873 	ret = __sendmsg(cinfo, &cmsg);
874 	unlock_comm(cinfo);
875 	return ret;
876 }
877 
878 static int new_disk_ack(struct mddev *mddev, bool ack)
879 {
880 	struct md_cluster_info *cinfo = mddev->cluster_info;
881 
882 	if (!test_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state)) {
883 		pr_warn("md-cluster(%s): Spurious cluster confirmation\n", mdname(mddev));
884 		return -EINVAL;
885 	}
886 
887 	if (ack)
888 		dlm_unlock_sync(cinfo->no_new_dev_lockres);
889 	complete(&cinfo->newdisk_completion);
890 	return 0;
891 }
892 
893 static int remove_disk(struct mddev *mddev, struct md_rdev *rdev)
894 {
895 	struct cluster_msg cmsg;
896 	struct md_cluster_info *cinfo = mddev->cluster_info;
897 	cmsg.type = REMOVE;
898 	cmsg.raid_slot = rdev->desc_nr;
899 	return __sendmsg(cinfo, &cmsg);
900 }
901 
902 static int gather_bitmaps(struct md_rdev *rdev)
903 {
904 	int sn, err;
905 	sector_t lo, hi;
906 	struct cluster_msg cmsg;
907 	struct mddev *mddev = rdev->mddev;
908 	struct md_cluster_info *cinfo = mddev->cluster_info;
909 
910 	cmsg.type = RE_ADD;
911 	cmsg.raid_slot = rdev->desc_nr;
912 	err = sendmsg(cinfo, &cmsg);
913 	if (err)
914 		goto out;
915 
916 	for (sn = 0; sn < mddev->bitmap_info.nodes; sn++) {
917 		if (sn == (cinfo->slot_number - 1))
918 			continue;
919 		err = bitmap_copy_from_slot(mddev, sn, &lo, &hi, false);
920 		if (err) {
921 			pr_warn("md-cluster: Could not gather bitmaps from slot %d", sn);
922 			goto out;
923 		}
924 		if ((hi > 0) && (lo < mddev->recovery_cp))
925 			mddev->recovery_cp = lo;
926 	}
927 out:
928 	return err;
929 }
930 
931 static struct md_cluster_operations cluster_ops = {
932 	.join   = join,
933 	.leave  = leave,
934 	.slot_number = slot_number,
935 	.resync_info_update = resync_info_update,
936 	.resync_start = resync_start,
937 	.resync_finish = resync_finish,
938 	.metadata_update_start = metadata_update_start,
939 	.metadata_update_finish = metadata_update_finish,
940 	.metadata_update_cancel = metadata_update_cancel,
941 	.area_resyncing = area_resyncing,
942 	.add_new_disk_start = add_new_disk_start,
943 	.add_new_disk_finish = add_new_disk_finish,
944 	.new_disk_ack = new_disk_ack,
945 	.remove_disk = remove_disk,
946 	.gather_bitmaps = gather_bitmaps,
947 };
948 
949 static int __init cluster_init(void)
950 {
951 	pr_warn("md-cluster: EXPERIMENTAL. Use with caution\n");
952 	pr_info("Registering Cluster MD functions\n");
953 	register_md_cluster_operations(&cluster_ops, THIS_MODULE);
954 	return 0;
955 }
956 
957 static void cluster_exit(void)
958 {
959 	unregister_md_cluster_operations();
960 }
961 
962 module_init(cluster_init);
963 module_exit(cluster_exit);
964 MODULE_LICENSE("GPL");
965 MODULE_DESCRIPTION("Clustering support for MD");
966