xref: /openbmc/linux/drivers/md/md-cluster.c (revision 9cfc5c90)
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 	int mode;
32 };
33 
34 struct suspend_info {
35 	int slot;
36 	sector_t lo;
37 	sector_t hi;
38 	struct list_head list;
39 };
40 
41 struct resync_info {
42 	__le64 lo;
43 	__le64 hi;
44 };
45 
46 /* md_cluster_info flags */
47 #define		MD_CLUSTER_WAITING_FOR_NEWDISK		1
48 #define		MD_CLUSTER_SUSPEND_READ_BALANCING	2
49 #define		MD_CLUSTER_BEGIN_JOIN_CLUSTER		3
50 
51 
52 struct md_cluster_info {
53 	/* dlm lock space and resources for clustered raid. */
54 	dlm_lockspace_t *lockspace;
55 	int slot_number;
56 	struct completion completion;
57 	struct dlm_lock_resource *bitmap_lockres;
58 	struct dlm_lock_resource *resync_lockres;
59 	struct list_head suspend_list;
60 	spinlock_t suspend_lock;
61 	struct md_thread *recovery_thread;
62 	unsigned long recovery_map;
63 	/* communication loc resources */
64 	struct dlm_lock_resource *ack_lockres;
65 	struct dlm_lock_resource *message_lockres;
66 	struct dlm_lock_resource *token_lockres;
67 	struct dlm_lock_resource *no_new_dev_lockres;
68 	struct md_thread *recv_thread;
69 	struct completion newdisk_completion;
70 	unsigned long state;
71 };
72 
73 enum msg_type {
74 	METADATA_UPDATED = 0,
75 	RESYNCING,
76 	NEWDISK,
77 	REMOVE,
78 	RE_ADD,
79 	BITMAP_NEEDS_SYNC,
80 };
81 
82 struct cluster_msg {
83 	__le32 type;
84 	__le32 slot;
85 	/* TODO: Unionize this for smaller footprint */
86 	__le64 low;
87 	__le64 high;
88 	char uuid[16];
89 	__le32 raid_slot;
90 };
91 
92 static void sync_ast(void *arg)
93 {
94 	struct dlm_lock_resource *res;
95 
96 	res = arg;
97 	complete(&res->completion);
98 }
99 
100 static int dlm_lock_sync(struct dlm_lock_resource *res, int mode)
101 {
102 	int ret = 0;
103 
104 	ret = dlm_lock(res->ls, mode, &res->lksb,
105 			res->flags, res->name, strlen(res->name),
106 			0, sync_ast, res, res->bast);
107 	if (ret)
108 		return ret;
109 	wait_for_completion(&res->completion);
110 	if (res->lksb.sb_status == 0)
111 		res->mode = mode;
112 	return res->lksb.sb_status;
113 }
114 
115 static int dlm_unlock_sync(struct dlm_lock_resource *res)
116 {
117 	return dlm_lock_sync(res, DLM_LOCK_NL);
118 }
119 
120 static struct dlm_lock_resource *lockres_init(struct mddev *mddev,
121 		char *name, void (*bastfn)(void *arg, int mode), int with_lvb)
122 {
123 	struct dlm_lock_resource *res = NULL;
124 	int ret, namelen;
125 	struct md_cluster_info *cinfo = mddev->cluster_info;
126 
127 	res = kzalloc(sizeof(struct dlm_lock_resource), GFP_KERNEL);
128 	if (!res)
129 		return NULL;
130 	init_completion(&res->completion);
131 	res->ls = cinfo->lockspace;
132 	res->mddev = mddev;
133 	res->mode = DLM_LOCK_IV;
134 	namelen = strlen(name);
135 	res->name = kzalloc(namelen + 1, GFP_KERNEL);
136 	if (!res->name) {
137 		pr_err("md-cluster: Unable to allocate resource name for resource %s\n", name);
138 		goto out_err;
139 	}
140 	strlcpy(res->name, name, namelen + 1);
141 	if (with_lvb) {
142 		res->lksb.sb_lvbptr = kzalloc(LVB_SIZE, GFP_KERNEL);
143 		if (!res->lksb.sb_lvbptr) {
144 			pr_err("md-cluster: Unable to allocate LVB for resource %s\n", name);
145 			goto out_err;
146 		}
147 		res->flags = DLM_LKF_VALBLK;
148 	}
149 
150 	if (bastfn)
151 		res->bast = bastfn;
152 
153 	res->flags |= DLM_LKF_EXPEDITE;
154 
155 	ret = dlm_lock_sync(res, DLM_LOCK_NL);
156 	if (ret) {
157 		pr_err("md-cluster: Unable to lock NL on new lock resource %s\n", name);
158 		goto out_err;
159 	}
160 	res->flags &= ~DLM_LKF_EXPEDITE;
161 	res->flags |= DLM_LKF_CONVERT;
162 
163 	return res;
164 out_err:
165 	kfree(res->lksb.sb_lvbptr);
166 	kfree(res->name);
167 	kfree(res);
168 	return NULL;
169 }
170 
171 static void lockres_free(struct dlm_lock_resource *res)
172 {
173 	int ret;
174 
175 	if (!res)
176 		return;
177 
178 	/* cancel a lock request or a conversion request that is blocked */
179 	res->flags |= DLM_LKF_CANCEL;
180 retry:
181 	ret = dlm_unlock(res->ls, res->lksb.sb_lkid, 0, &res->lksb, res);
182 	if (unlikely(ret != 0)) {
183 		pr_info("%s: failed to unlock %s return %d\n", __func__, res->name, ret);
184 
185 		/* if a lock conversion is cancelled, then the lock is put
186 		 * back to grant queue, need to ensure it is unlocked */
187 		if (ret == -DLM_ECANCEL)
188 			goto retry;
189 	}
190 	res->flags &= ~DLM_LKF_CANCEL;
191 	wait_for_completion(&res->completion);
192 
193 	kfree(res->name);
194 	kfree(res->lksb.sb_lvbptr);
195 	kfree(res);
196 }
197 
198 static void add_resync_info(struct dlm_lock_resource *lockres,
199 			    sector_t lo, sector_t hi)
200 {
201 	struct resync_info *ri;
202 
203 	ri = (struct resync_info *)lockres->lksb.sb_lvbptr;
204 	ri->lo = cpu_to_le64(lo);
205 	ri->hi = cpu_to_le64(hi);
206 }
207 
208 static struct suspend_info *read_resync_info(struct mddev *mddev, struct dlm_lock_resource *lockres)
209 {
210 	struct resync_info ri;
211 	struct suspend_info *s = NULL;
212 	sector_t hi = 0;
213 
214 	dlm_lock_sync(lockres, DLM_LOCK_CR);
215 	memcpy(&ri, lockres->lksb.sb_lvbptr, sizeof(struct resync_info));
216 	hi = le64_to_cpu(ri.hi);
217 	if (hi > 0) {
218 		s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL);
219 		if (!s)
220 			goto out;
221 		s->hi = hi;
222 		s->lo = le64_to_cpu(ri.lo);
223 	}
224 	dlm_unlock_sync(lockres);
225 out:
226 	return s;
227 }
228 
229 static void recover_bitmaps(struct md_thread *thread)
230 {
231 	struct mddev *mddev = thread->mddev;
232 	struct md_cluster_info *cinfo = mddev->cluster_info;
233 	struct dlm_lock_resource *bm_lockres;
234 	char str[64];
235 	int slot, ret;
236 	struct suspend_info *s, *tmp;
237 	sector_t lo, hi;
238 
239 	while (cinfo->recovery_map) {
240 		slot = fls64((u64)cinfo->recovery_map) - 1;
241 
242 		/* Clear suspend_area associated with the bitmap */
243 		spin_lock_irq(&cinfo->suspend_lock);
244 		list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list)
245 			if (slot == s->slot) {
246 				list_del(&s->list);
247 				kfree(s);
248 			}
249 		spin_unlock_irq(&cinfo->suspend_lock);
250 
251 		snprintf(str, 64, "bitmap%04d", slot);
252 		bm_lockres = lockres_init(mddev, str, NULL, 1);
253 		if (!bm_lockres) {
254 			pr_err("md-cluster: Cannot initialize bitmaps\n");
255 			goto clear_bit;
256 		}
257 
258 		ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
259 		if (ret) {
260 			pr_err("md-cluster: Could not DLM lock %s: %d\n",
261 					str, ret);
262 			goto clear_bit;
263 		}
264 		ret = bitmap_copy_from_slot(mddev, slot, &lo, &hi, true);
265 		if (ret) {
266 			pr_err("md-cluster: Could not copy data from bitmap %d\n", slot);
267 			goto dlm_unlock;
268 		}
269 		if (hi > 0) {
270 			/* TODO:Wait for current resync to get over */
271 			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
272 			if (lo < mddev->recovery_cp)
273 				mddev->recovery_cp = lo;
274 			md_check_recovery(mddev);
275 		}
276 dlm_unlock:
277 		dlm_unlock_sync(bm_lockres);
278 clear_bit:
279 		clear_bit(slot, &cinfo->recovery_map);
280 	}
281 }
282 
283 static void recover_prep(void *arg)
284 {
285 	struct mddev *mddev = arg;
286 	struct md_cluster_info *cinfo = mddev->cluster_info;
287 	set_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state);
288 }
289 
290 static void __recover_slot(struct mddev *mddev, int slot)
291 {
292 	struct md_cluster_info *cinfo = mddev->cluster_info;
293 
294 	set_bit(slot, &cinfo->recovery_map);
295 	if (!cinfo->recovery_thread) {
296 		cinfo->recovery_thread = md_register_thread(recover_bitmaps,
297 				mddev, "recover");
298 		if (!cinfo->recovery_thread) {
299 			pr_warn("md-cluster: Could not create recovery thread\n");
300 			return;
301 		}
302 	}
303 	md_wakeup_thread(cinfo->recovery_thread);
304 }
305 
306 static void recover_slot(void *arg, struct dlm_slot *slot)
307 {
308 	struct mddev *mddev = arg;
309 	struct md_cluster_info *cinfo = mddev->cluster_info;
310 
311 	pr_info("md-cluster: %s Node %d/%d down. My slot: %d. Initiating recovery.\n",
312 			mddev->bitmap_info.cluster_name,
313 			slot->nodeid, slot->slot,
314 			cinfo->slot_number);
315 	/* deduct one since dlm slot starts from one while the num of
316 	 * cluster-md begins with 0 */
317 	__recover_slot(mddev, slot->slot - 1);
318 }
319 
320 static void recover_done(void *arg, struct dlm_slot *slots,
321 		int num_slots, int our_slot,
322 		uint32_t generation)
323 {
324 	struct mddev *mddev = arg;
325 	struct md_cluster_info *cinfo = mddev->cluster_info;
326 
327 	cinfo->slot_number = our_slot;
328 	/* completion is only need to be complete when node join cluster,
329 	 * it doesn't need to run during another node's failure */
330 	if (test_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state)) {
331 		complete(&cinfo->completion);
332 		clear_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state);
333 	}
334 	clear_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state);
335 }
336 
337 /* the ops is called when node join the cluster, and do lock recovery
338  * if node failure occurs */
339 static const struct dlm_lockspace_ops md_ls_ops = {
340 	.recover_prep = recover_prep,
341 	.recover_slot = recover_slot,
342 	.recover_done = recover_done,
343 };
344 
345 /*
346  * The BAST function for the ack lock resource
347  * This function wakes up the receive thread in
348  * order to receive and process the message.
349  */
350 static void ack_bast(void *arg, int mode)
351 {
352 	struct dlm_lock_resource *res = arg;
353 	struct md_cluster_info *cinfo = res->mddev->cluster_info;
354 
355 	if (mode == DLM_LOCK_EX)
356 		md_wakeup_thread(cinfo->recv_thread);
357 }
358 
359 static void __remove_suspend_info(struct md_cluster_info *cinfo, int slot)
360 {
361 	struct suspend_info *s, *tmp;
362 
363 	list_for_each_entry_safe(s, tmp, &cinfo->suspend_list, list)
364 		if (slot == s->slot) {
365 			list_del(&s->list);
366 			kfree(s);
367 			break;
368 		}
369 }
370 
371 static void remove_suspend_info(struct mddev *mddev, int slot)
372 {
373 	struct md_cluster_info *cinfo = mddev->cluster_info;
374 	spin_lock_irq(&cinfo->suspend_lock);
375 	__remove_suspend_info(cinfo, slot);
376 	spin_unlock_irq(&cinfo->suspend_lock);
377 	mddev->pers->quiesce(mddev, 2);
378 }
379 
380 
381 static void process_suspend_info(struct mddev *mddev,
382 		int slot, sector_t lo, sector_t hi)
383 {
384 	struct md_cluster_info *cinfo = mddev->cluster_info;
385 	struct suspend_info *s;
386 
387 	if (!hi) {
388 		remove_suspend_info(mddev, slot);
389 		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
390 		md_wakeup_thread(mddev->thread);
391 		return;
392 	}
393 	s = kzalloc(sizeof(struct suspend_info), GFP_KERNEL);
394 	if (!s)
395 		return;
396 	s->slot = slot;
397 	s->lo = lo;
398 	s->hi = hi;
399 	mddev->pers->quiesce(mddev, 1);
400 	mddev->pers->quiesce(mddev, 0);
401 	spin_lock_irq(&cinfo->suspend_lock);
402 	/* Remove existing entry (if exists) before adding */
403 	__remove_suspend_info(cinfo, slot);
404 	list_add(&s->list, &cinfo->suspend_list);
405 	spin_unlock_irq(&cinfo->suspend_lock);
406 	mddev->pers->quiesce(mddev, 2);
407 }
408 
409 static void process_add_new_disk(struct mddev *mddev, struct cluster_msg *cmsg)
410 {
411 	char disk_uuid[64];
412 	struct md_cluster_info *cinfo = mddev->cluster_info;
413 	char event_name[] = "EVENT=ADD_DEVICE";
414 	char raid_slot[16];
415 	char *envp[] = {event_name, disk_uuid, raid_slot, NULL};
416 	int len;
417 
418 	len = snprintf(disk_uuid, 64, "DEVICE_UUID=");
419 	sprintf(disk_uuid + len, "%pU", cmsg->uuid);
420 	snprintf(raid_slot, 16, "RAID_DISK=%d", le32_to_cpu(cmsg->raid_slot));
421 	pr_info("%s:%d Sending kobject change with %s and %s\n", __func__, __LINE__, disk_uuid, raid_slot);
422 	init_completion(&cinfo->newdisk_completion);
423 	set_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
424 	kobject_uevent_env(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE, envp);
425 	wait_for_completion_timeout(&cinfo->newdisk_completion,
426 			NEW_DEV_TIMEOUT);
427 	clear_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state);
428 }
429 
430 
431 static void process_metadata_update(struct mddev *mddev, struct cluster_msg *msg)
432 {
433 	struct md_cluster_info *cinfo = mddev->cluster_info;
434 	md_reload_sb(mddev, le32_to_cpu(msg->raid_slot));
435 	dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
436 }
437 
438 static void process_remove_disk(struct mddev *mddev, struct cluster_msg *msg)
439 {
440 	struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev,
441 						   le32_to_cpu(msg->raid_slot));
442 
443 	if (rdev)
444 		md_kick_rdev_from_array(rdev);
445 	else
446 		pr_warn("%s: %d Could not find disk(%d) to REMOVE\n",
447 			__func__, __LINE__, le32_to_cpu(msg->raid_slot));
448 }
449 
450 static void process_readd_disk(struct mddev *mddev, struct cluster_msg *msg)
451 {
452 	struct md_rdev *rdev = md_find_rdev_nr_rcu(mddev,
453 						   le32_to_cpu(msg->raid_slot));
454 
455 	if (rdev && test_bit(Faulty, &rdev->flags))
456 		clear_bit(Faulty, &rdev->flags);
457 	else
458 		pr_warn("%s: %d Could not find disk(%d) which is faulty",
459 			__func__, __LINE__, le32_to_cpu(msg->raid_slot));
460 }
461 
462 static void process_recvd_msg(struct mddev *mddev, struct cluster_msg *msg)
463 {
464 	if (WARN(mddev->cluster_info->slot_number - 1 == le32_to_cpu(msg->slot),
465 		"node %d received it's own msg\n", le32_to_cpu(msg->slot)))
466 		return;
467 	switch (le32_to_cpu(msg->type)) {
468 	case METADATA_UPDATED:
469 		process_metadata_update(mddev, msg);
470 		break;
471 	case RESYNCING:
472 		process_suspend_info(mddev, le32_to_cpu(msg->slot),
473 				     le64_to_cpu(msg->low),
474 				     le64_to_cpu(msg->high));
475 		break;
476 	case NEWDISK:
477 		process_add_new_disk(mddev, msg);
478 		break;
479 	case REMOVE:
480 		process_remove_disk(mddev, msg);
481 		break;
482 	case RE_ADD:
483 		process_readd_disk(mddev, msg);
484 		break;
485 	case BITMAP_NEEDS_SYNC:
486 		__recover_slot(mddev, le32_to_cpu(msg->slot));
487 		break;
488 	default:
489 		pr_warn("%s:%d Received unknown message from %d\n",
490 			__func__, __LINE__, msg->slot);
491 	}
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;
503 	int ret;
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*/
516 	ret = dlm_unlock_sync(ack_lockres);
517 	if (unlikely(ret != 0))
518 		pr_info("unlock ack failed return %d\n", ret);
519 	/*up-convert to PR on message_lockres*/
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);
523 	/*get CR on ack_lockres again*/
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);
527 	/*release CR on message_lockres*/
528 	ret = dlm_unlock_sync(message_lockres);
529 	if (unlikely(ret != 0))
530 		pr_info("unlock msg failed return %d\n", ret);
531 }
532 
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  * If called again, and the TOKEN lock is alread in EX mode
537  * return success. However, care must be taken that unlock_comm()
538  * is called only once.
539  */
540 static int lock_comm(struct md_cluster_info *cinfo)
541 {
542 	int error;
543 
544 	if (cinfo->token_lockres->mode == DLM_LOCK_EX)
545 		return 0;
546 
547 	error = dlm_lock_sync(cinfo->token_lockres, DLM_LOCK_EX);
548 	if (error)
549 		pr_err("md-cluster(%s:%d): failed to get EX on TOKEN (%d)\n",
550 				__func__, __LINE__, error);
551 	return error;
552 }
553 
554 static void unlock_comm(struct md_cluster_info *cinfo)
555 {
556 	WARN_ON(cinfo->token_lockres->mode != DLM_LOCK_EX);
557 	dlm_unlock_sync(cinfo->token_lockres);
558 }
559 
560 /* __sendmsg()
561  * This function performs the actual sending of the message. This function is
562  * usually called after performing the encompassing operation
563  * The function:
564  * 1. Grabs the message lockresource in EX mode
565  * 2. Copies the message to the message LVB
566  * 3. Downconverts message lockresource to CW
567  * 4. Upconverts ack lock resource from CR to EX. This forces the BAST on other nodes
568  *    and the other nodes read the message. The thread will wait here until all other
569  *    nodes have released ack lock resource.
570  * 5. Downconvert ack lockresource to CR
571  */
572 static int __sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
573 {
574 	int error;
575 	int slot = cinfo->slot_number - 1;
576 
577 	cmsg->slot = cpu_to_le32(slot);
578 	/*get EX on Message*/
579 	error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_EX);
580 	if (error) {
581 		pr_err("md-cluster: failed to get EX on MESSAGE (%d)\n", error);
582 		goto failed_message;
583 	}
584 
585 	memcpy(cinfo->message_lockres->lksb.sb_lvbptr, (void *)cmsg,
586 			sizeof(struct cluster_msg));
587 	/*down-convert EX to CW on Message*/
588 	error = dlm_lock_sync(cinfo->message_lockres, DLM_LOCK_CW);
589 	if (error) {
590 		pr_err("md-cluster: failed to convert EX to CW on MESSAGE(%d)\n",
591 				error);
592 		goto failed_ack;
593 	}
594 
595 	/*up-convert CR to EX on Ack*/
596 	error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_EX);
597 	if (error) {
598 		pr_err("md-cluster: failed to convert CR to EX on ACK(%d)\n",
599 				error);
600 		goto failed_ack;
601 	}
602 
603 	/*down-convert EX to CR on Ack*/
604 	error = dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR);
605 	if (error) {
606 		pr_err("md-cluster: failed to convert EX to CR on ACK(%d)\n",
607 				error);
608 		goto failed_ack;
609 	}
610 
611 failed_ack:
612 	error = dlm_unlock_sync(cinfo->message_lockres);
613 	if (unlikely(error != 0)) {
614 		pr_err("md-cluster: failed convert to NL on MESSAGE(%d)\n",
615 			error);
616 		/* in case the message can't be released due to some reason */
617 		goto failed_ack;
618 	}
619 failed_message:
620 	return error;
621 }
622 
623 static int sendmsg(struct md_cluster_info *cinfo, struct cluster_msg *cmsg)
624 {
625 	int ret;
626 
627 	lock_comm(cinfo);
628 	ret = __sendmsg(cinfo, cmsg);
629 	unlock_comm(cinfo);
630 	return ret;
631 }
632 
633 static int gather_all_resync_info(struct mddev *mddev, int total_slots)
634 {
635 	struct md_cluster_info *cinfo = mddev->cluster_info;
636 	int i, ret = 0;
637 	struct dlm_lock_resource *bm_lockres;
638 	struct suspend_info *s;
639 	char str[64];
640 	sector_t lo, hi;
641 
642 
643 	for (i = 0; i < total_slots; i++) {
644 		memset(str, '\0', 64);
645 		snprintf(str, 64, "bitmap%04d", i);
646 		bm_lockres = lockres_init(mddev, str, NULL, 1);
647 		if (!bm_lockres)
648 			return -ENOMEM;
649 		if (i == (cinfo->slot_number - 1))
650 			continue;
651 
652 		bm_lockres->flags |= DLM_LKF_NOQUEUE;
653 		ret = dlm_lock_sync(bm_lockres, DLM_LOCK_PW);
654 		if (ret == -EAGAIN) {
655 			memset(bm_lockres->lksb.sb_lvbptr, '\0', LVB_SIZE);
656 			s = read_resync_info(mddev, bm_lockres);
657 			if (s) {
658 				pr_info("%s:%d Resync[%llu..%llu] in progress on %d\n",
659 						__func__, __LINE__,
660 						(unsigned long long) s->lo,
661 						(unsigned long long) s->hi, i);
662 				spin_lock_irq(&cinfo->suspend_lock);
663 				s->slot = i;
664 				list_add(&s->list, &cinfo->suspend_list);
665 				spin_unlock_irq(&cinfo->suspend_lock);
666 			}
667 			ret = 0;
668 			lockres_free(bm_lockres);
669 			continue;
670 		}
671 		if (ret) {
672 			lockres_free(bm_lockres);
673 			goto out;
674 		}
675 
676 		/* Read the disk bitmap sb and check if it needs recovery */
677 		ret = bitmap_copy_from_slot(mddev, i, &lo, &hi, false);
678 		if (ret) {
679 			pr_warn("md-cluster: Could not gather bitmaps from slot %d", i);
680 			lockres_free(bm_lockres);
681 			continue;
682 		}
683 		if ((hi > 0) && (lo < mddev->recovery_cp)) {
684 			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
685 			mddev->recovery_cp = lo;
686 			md_check_recovery(mddev);
687 		}
688 
689 		dlm_unlock_sync(bm_lockres);
690 		lockres_free(bm_lockres);
691 	}
692 out:
693 	return ret;
694 }
695 
696 static int join(struct mddev *mddev, int nodes)
697 {
698 	struct md_cluster_info *cinfo;
699 	int ret, ops_rv;
700 	char str[64];
701 
702 	cinfo = kzalloc(sizeof(struct md_cluster_info), GFP_KERNEL);
703 	if (!cinfo)
704 		return -ENOMEM;
705 
706 	INIT_LIST_HEAD(&cinfo->suspend_list);
707 	spin_lock_init(&cinfo->suspend_lock);
708 	init_completion(&cinfo->completion);
709 	set_bit(MD_CLUSTER_BEGIN_JOIN_CLUSTER, &cinfo->state);
710 
711 	mddev->cluster_info = cinfo;
712 
713 	memset(str, 0, 64);
714 	sprintf(str, "%pU", mddev->uuid);
715 	ret = dlm_new_lockspace(str, mddev->bitmap_info.cluster_name,
716 				DLM_LSFL_FS, LVB_SIZE,
717 				&md_ls_ops, mddev, &ops_rv, &cinfo->lockspace);
718 	if (ret)
719 		goto err;
720 	wait_for_completion(&cinfo->completion);
721 	if (nodes < cinfo->slot_number) {
722 		pr_err("md-cluster: Slot allotted(%d) is greater than available slots(%d).",
723 			cinfo->slot_number, nodes);
724 		ret = -ERANGE;
725 		goto err;
726 	}
727 	/* Initiate the communication resources */
728 	ret = -ENOMEM;
729 	cinfo->recv_thread = md_register_thread(recv_daemon, mddev, "cluster_recv");
730 	if (!cinfo->recv_thread) {
731 		pr_err("md-cluster: cannot allocate memory for recv_thread!\n");
732 		goto err;
733 	}
734 	cinfo->message_lockres = lockres_init(mddev, "message", NULL, 1);
735 	if (!cinfo->message_lockres)
736 		goto err;
737 	cinfo->token_lockres = lockres_init(mddev, "token", NULL, 0);
738 	if (!cinfo->token_lockres)
739 		goto err;
740 	cinfo->ack_lockres = lockres_init(mddev, "ack", ack_bast, 0);
741 	if (!cinfo->ack_lockres)
742 		goto err;
743 	cinfo->no_new_dev_lockres = lockres_init(mddev, "no-new-dev", NULL, 0);
744 	if (!cinfo->no_new_dev_lockres)
745 		goto err;
746 
747 	/* get sync CR lock on ACK. */
748 	if (dlm_lock_sync(cinfo->ack_lockres, DLM_LOCK_CR))
749 		pr_err("md-cluster: failed to get a sync CR lock on ACK!(%d)\n",
750 				ret);
751 	/* get sync CR lock on no-new-dev. */
752 	if (dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR))
753 		pr_err("md-cluster: failed to get a sync CR lock on no-new-dev!(%d)\n", ret);
754 
755 
756 	pr_info("md-cluster: Joined cluster %s slot %d\n", str, cinfo->slot_number);
757 	snprintf(str, 64, "bitmap%04d", cinfo->slot_number - 1);
758 	cinfo->bitmap_lockres = lockres_init(mddev, str, NULL, 1);
759 	if (!cinfo->bitmap_lockres)
760 		goto err;
761 	if (dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW)) {
762 		pr_err("Failed to get bitmap lock\n");
763 		ret = -EINVAL;
764 		goto err;
765 	}
766 
767 	cinfo->resync_lockres = lockres_init(mddev, "resync", NULL, 0);
768 	if (!cinfo->resync_lockres)
769 		goto err;
770 
771 	ret = gather_all_resync_info(mddev, nodes);
772 	if (ret)
773 		goto err;
774 
775 	return 0;
776 err:
777 	lockres_free(cinfo->message_lockres);
778 	lockres_free(cinfo->token_lockres);
779 	lockres_free(cinfo->ack_lockres);
780 	lockres_free(cinfo->no_new_dev_lockres);
781 	lockres_free(cinfo->resync_lockres);
782 	lockres_free(cinfo->bitmap_lockres);
783 	if (cinfo->lockspace)
784 		dlm_release_lockspace(cinfo->lockspace, 2);
785 	mddev->cluster_info = NULL;
786 	kfree(cinfo);
787 	return ret;
788 }
789 
790 static void resync_bitmap(struct mddev *mddev)
791 {
792 	struct md_cluster_info *cinfo = mddev->cluster_info;
793 	struct cluster_msg cmsg = {0};
794 	int err;
795 
796 	cmsg.type = cpu_to_le32(BITMAP_NEEDS_SYNC);
797 	err = sendmsg(cinfo, &cmsg);
798 	if (err)
799 		pr_err("%s:%d: failed to send BITMAP_NEEDS_SYNC message (%d)\n",
800 			__func__, __LINE__, err);
801 }
802 
803 static int leave(struct mddev *mddev)
804 {
805 	struct md_cluster_info *cinfo = mddev->cluster_info;
806 
807 	if (!cinfo)
808 		return 0;
809 
810 	/* BITMAP_NEEDS_SYNC message should be sent when node
811 	 * is leaving the cluster with dirty bitmap, also we
812 	 * can only deliver it when dlm connection is available */
813 	if (cinfo->slot_number > 0 && mddev->recovery_cp != MaxSector)
814 		resync_bitmap(mddev);
815 
816 	md_unregister_thread(&cinfo->recovery_thread);
817 	md_unregister_thread(&cinfo->recv_thread);
818 	lockres_free(cinfo->message_lockres);
819 	lockres_free(cinfo->token_lockres);
820 	lockres_free(cinfo->ack_lockres);
821 	lockres_free(cinfo->no_new_dev_lockres);
822 	lockres_free(cinfo->bitmap_lockres);
823 	dlm_release_lockspace(cinfo->lockspace, 2);
824 	return 0;
825 }
826 
827 /* slot_number(): Returns the MD slot number to use
828  * DLM starts the slot numbers from 1, wheras cluster-md
829  * wants the number to be from zero, so we deduct one
830  */
831 static int slot_number(struct mddev *mddev)
832 {
833 	struct md_cluster_info *cinfo = mddev->cluster_info;
834 
835 	return cinfo->slot_number - 1;
836 }
837 
838 static int metadata_update_start(struct mddev *mddev)
839 {
840 	return lock_comm(mddev->cluster_info);
841 }
842 
843 static int metadata_update_finish(struct mddev *mddev)
844 {
845 	struct md_cluster_info *cinfo = mddev->cluster_info;
846 	struct cluster_msg cmsg;
847 	struct md_rdev *rdev;
848 	int ret = 0;
849 	int raid_slot = -1;
850 
851 	memset(&cmsg, 0, sizeof(cmsg));
852 	cmsg.type = cpu_to_le32(METADATA_UPDATED);
853 	/* Pick up a good active device number to send.
854 	 */
855 	rdev_for_each(rdev, mddev)
856 		if (rdev->raid_disk > -1 && !test_bit(Faulty, &rdev->flags)) {
857 			raid_slot = rdev->desc_nr;
858 			break;
859 		}
860 	if (raid_slot >= 0) {
861 		cmsg.raid_slot = cpu_to_le32(raid_slot);
862 		ret = __sendmsg(cinfo, &cmsg);
863 	} else
864 		pr_warn("md-cluster: No good device id found to send\n");
865 	unlock_comm(cinfo);
866 	return ret;
867 }
868 
869 static void metadata_update_cancel(struct mddev *mddev)
870 {
871 	struct md_cluster_info *cinfo = mddev->cluster_info;
872 	unlock_comm(cinfo);
873 }
874 
875 static int resync_start(struct mddev *mddev)
876 {
877 	struct md_cluster_info *cinfo = mddev->cluster_info;
878 	cinfo->resync_lockres->flags |= DLM_LKF_NOQUEUE;
879 	return dlm_lock_sync(cinfo->resync_lockres, DLM_LOCK_EX);
880 }
881 
882 static int resync_info_update(struct mddev *mddev, sector_t lo, sector_t hi)
883 {
884 	struct md_cluster_info *cinfo = mddev->cluster_info;
885 	struct cluster_msg cmsg = {0};
886 
887 	add_resync_info(cinfo->bitmap_lockres, lo, hi);
888 	/* Re-acquire the lock to refresh LVB */
889 	dlm_lock_sync(cinfo->bitmap_lockres, DLM_LOCK_PW);
890 	cmsg.type = cpu_to_le32(RESYNCING);
891 	cmsg.low = cpu_to_le64(lo);
892 	cmsg.high = cpu_to_le64(hi);
893 
894 	return sendmsg(cinfo, &cmsg);
895 }
896 
897 static int resync_finish(struct mddev *mddev)
898 {
899 	struct md_cluster_info *cinfo = mddev->cluster_info;
900 	cinfo->resync_lockres->flags &= ~DLM_LKF_NOQUEUE;
901 	dlm_unlock_sync(cinfo->resync_lockres);
902 	return resync_info_update(mddev, 0, 0);
903 }
904 
905 static int area_resyncing(struct mddev *mddev, int direction,
906 		sector_t lo, sector_t hi)
907 {
908 	struct md_cluster_info *cinfo = mddev->cluster_info;
909 	int ret = 0;
910 	struct suspend_info *s;
911 
912 	if ((direction == READ) &&
913 		test_bit(MD_CLUSTER_SUSPEND_READ_BALANCING, &cinfo->state))
914 		return 1;
915 
916 	spin_lock_irq(&cinfo->suspend_lock);
917 	if (list_empty(&cinfo->suspend_list))
918 		goto out;
919 	list_for_each_entry(s, &cinfo->suspend_list, list)
920 		if (hi > s->lo && lo < s->hi) {
921 			ret = 1;
922 			break;
923 		}
924 out:
925 	spin_unlock_irq(&cinfo->suspend_lock);
926 	return ret;
927 }
928 
929 /* add_new_disk() - initiates a disk add
930  * However, if this fails before writing md_update_sb(),
931  * add_new_disk_cancel() must be called to release token lock
932  */
933 static int add_new_disk(struct mddev *mddev, struct md_rdev *rdev)
934 {
935 	struct md_cluster_info *cinfo = mddev->cluster_info;
936 	struct cluster_msg cmsg;
937 	int ret = 0;
938 	struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
939 	char *uuid = sb->device_uuid;
940 
941 	memset(&cmsg, 0, sizeof(cmsg));
942 	cmsg.type = cpu_to_le32(NEWDISK);
943 	memcpy(cmsg.uuid, uuid, 16);
944 	cmsg.raid_slot = cpu_to_le32(rdev->desc_nr);
945 	lock_comm(cinfo);
946 	ret = __sendmsg(cinfo, &cmsg);
947 	if (ret)
948 		return ret;
949 	cinfo->no_new_dev_lockres->flags |= DLM_LKF_NOQUEUE;
950 	ret = dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_EX);
951 	cinfo->no_new_dev_lockres->flags &= ~DLM_LKF_NOQUEUE;
952 	/* Some node does not "see" the device */
953 	if (ret == -EAGAIN)
954 		ret = -ENOENT;
955 	if (ret)
956 		unlock_comm(cinfo);
957 	else
958 		dlm_lock_sync(cinfo->no_new_dev_lockres, DLM_LOCK_CR);
959 	return ret;
960 }
961 
962 static void add_new_disk_cancel(struct mddev *mddev)
963 {
964 	struct md_cluster_info *cinfo = mddev->cluster_info;
965 	unlock_comm(cinfo);
966 }
967 
968 static int new_disk_ack(struct mddev *mddev, bool ack)
969 {
970 	struct md_cluster_info *cinfo = mddev->cluster_info;
971 
972 	if (!test_bit(MD_CLUSTER_WAITING_FOR_NEWDISK, &cinfo->state)) {
973 		pr_warn("md-cluster(%s): Spurious cluster confirmation\n", mdname(mddev));
974 		return -EINVAL;
975 	}
976 
977 	if (ack)
978 		dlm_unlock_sync(cinfo->no_new_dev_lockres);
979 	complete(&cinfo->newdisk_completion);
980 	return 0;
981 }
982 
983 static int remove_disk(struct mddev *mddev, struct md_rdev *rdev)
984 {
985 	struct cluster_msg cmsg = {0};
986 	struct md_cluster_info *cinfo = mddev->cluster_info;
987 	cmsg.type = cpu_to_le32(REMOVE);
988 	cmsg.raid_slot = cpu_to_le32(rdev->desc_nr);
989 	return __sendmsg(cinfo, &cmsg);
990 }
991 
992 static int gather_bitmaps(struct md_rdev *rdev)
993 {
994 	int sn, err;
995 	sector_t lo, hi;
996 	struct cluster_msg cmsg = {0};
997 	struct mddev *mddev = rdev->mddev;
998 	struct md_cluster_info *cinfo = mddev->cluster_info;
999 
1000 	cmsg.type = cpu_to_le32(RE_ADD);
1001 	cmsg.raid_slot = cpu_to_le32(rdev->desc_nr);
1002 	err = sendmsg(cinfo, &cmsg);
1003 	if (err)
1004 		goto out;
1005 
1006 	for (sn = 0; sn < mddev->bitmap_info.nodes; sn++) {
1007 		if (sn == (cinfo->slot_number - 1))
1008 			continue;
1009 		err = bitmap_copy_from_slot(mddev, sn, &lo, &hi, false);
1010 		if (err) {
1011 			pr_warn("md-cluster: Could not gather bitmaps from slot %d", sn);
1012 			goto out;
1013 		}
1014 		if ((hi > 0) && (lo < mddev->recovery_cp))
1015 			mddev->recovery_cp = lo;
1016 	}
1017 out:
1018 	return err;
1019 }
1020 
1021 static struct md_cluster_operations cluster_ops = {
1022 	.join   = join,
1023 	.leave  = leave,
1024 	.slot_number = slot_number,
1025 	.resync_start = resync_start,
1026 	.resync_finish = resync_finish,
1027 	.resync_info_update = resync_info_update,
1028 	.metadata_update_start = metadata_update_start,
1029 	.metadata_update_finish = metadata_update_finish,
1030 	.metadata_update_cancel = metadata_update_cancel,
1031 	.area_resyncing = area_resyncing,
1032 	.add_new_disk = add_new_disk,
1033 	.add_new_disk_cancel = add_new_disk_cancel,
1034 	.new_disk_ack = new_disk_ack,
1035 	.remove_disk = remove_disk,
1036 	.gather_bitmaps = gather_bitmaps,
1037 };
1038 
1039 static int __init cluster_init(void)
1040 {
1041 	pr_warn("md-cluster: EXPERIMENTAL. Use with caution\n");
1042 	pr_info("Registering Cluster MD functions\n");
1043 	register_md_cluster_operations(&cluster_ops, THIS_MODULE);
1044 	return 0;
1045 }
1046 
1047 static void cluster_exit(void)
1048 {
1049 	unregister_md_cluster_operations();
1050 }
1051 
1052 module_init(cluster_init);
1053 module_exit(cluster_exit);
1054 MODULE_AUTHOR("SUSE");
1055 MODULE_LICENSE("GPL");
1056 MODULE_DESCRIPTION("Clustering support for MD");
1057