xref: /linux/fs/ocfs2/cluster/heartbeat.c (revision 1517d90cfafe0f95fd7863d04e1596f7beb7dfa8)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* -*- mode: c; c-basic-offset: 8; -*-
3  * vim: noexpandtab sw=8 ts=8 sts=0:
4  *
5  * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
6  */
7 
8 #include <linux/kernel.h>
9 #include <linux/sched.h>
10 #include <linux/jiffies.h>
11 #include <linux/module.h>
12 #include <linux/fs.h>
13 #include <linux/bio.h>
14 #include <linux/blkdev.h>
15 #include <linux/delay.h>
16 #include <linux/file.h>
17 #include <linux/kthread.h>
18 #include <linux/configfs.h>
19 #include <linux/random.h>
20 #include <linux/crc32.h>
21 #include <linux/time.h>
22 #include <linux/debugfs.h>
23 #include <linux/slab.h>
24 #include <linux/bitmap.h>
25 #include <linux/ktime.h>
26 #include "heartbeat.h"
27 #include "tcp.h"
28 #include "nodemanager.h"
29 #include "quorum.h"
30 
31 #include "masklog.h"
32 
33 
34 /*
35  * The first heartbeat pass had one global thread that would serialize all hb
36  * callback calls.  This global serializing sem should only be removed once
37  * we've made sure that all callees can deal with being called concurrently
38  * from multiple hb region threads.
39  */
40 static DECLARE_RWSEM(o2hb_callback_sem);
41 
42 /*
43  * multiple hb threads are watching multiple regions.  A node is live
44  * whenever any of the threads sees activity from the node in its region.
45  */
46 static DEFINE_SPINLOCK(o2hb_live_lock);
47 static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
48 static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
49 static LIST_HEAD(o2hb_node_events);
50 static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
51 
52 /*
53  * In global heartbeat, we maintain a series of region bitmaps.
54  * 	- o2hb_region_bitmap allows us to limit the region number to max region.
55  * 	- o2hb_live_region_bitmap tracks live regions (seen steady iterations).
56  * 	- o2hb_quorum_region_bitmap tracks live regions that have seen all nodes
57  * 		heartbeat on it.
58  * 	- o2hb_failed_region_bitmap tracks the regions that have seen io timeouts.
59  */
60 static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
61 static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
62 static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
63 static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
64 
65 #define O2HB_DB_TYPE_LIVENODES		0
66 #define O2HB_DB_TYPE_LIVEREGIONS	1
67 #define O2HB_DB_TYPE_QUORUMREGIONS	2
68 #define O2HB_DB_TYPE_FAILEDREGIONS	3
69 #define O2HB_DB_TYPE_REGION_LIVENODES	4
70 #define O2HB_DB_TYPE_REGION_NUMBER	5
71 #define O2HB_DB_TYPE_REGION_ELAPSED_TIME	6
72 #define O2HB_DB_TYPE_REGION_PINNED	7
73 struct o2hb_debug_buf {
74 	int db_type;
75 	int db_size;
76 	int db_len;
77 	void *db_data;
78 };
79 
80 static struct o2hb_debug_buf *o2hb_db_livenodes;
81 static struct o2hb_debug_buf *o2hb_db_liveregions;
82 static struct o2hb_debug_buf *o2hb_db_quorumregions;
83 static struct o2hb_debug_buf *o2hb_db_failedregions;
84 
85 #define O2HB_DEBUG_DIR			"o2hb"
86 #define O2HB_DEBUG_LIVENODES		"livenodes"
87 #define O2HB_DEBUG_LIVEREGIONS		"live_regions"
88 #define O2HB_DEBUG_QUORUMREGIONS	"quorum_regions"
89 #define O2HB_DEBUG_FAILEDREGIONS	"failed_regions"
90 #define O2HB_DEBUG_REGION_NUMBER	"num"
91 #define O2HB_DEBUG_REGION_ELAPSED_TIME	"elapsed_time_in_ms"
92 #define O2HB_DEBUG_REGION_PINNED	"pinned"
93 
94 static struct dentry *o2hb_debug_dir;
95 
96 static LIST_HEAD(o2hb_all_regions);
97 
98 static struct o2hb_callback {
99 	struct list_head list;
100 } o2hb_callbacks[O2HB_NUM_CB];
101 
102 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
103 
104 #define O2HB_DEFAULT_BLOCK_BITS       9
105 
106 enum o2hb_heartbeat_modes {
107 	O2HB_HEARTBEAT_LOCAL		= 0,
108 	O2HB_HEARTBEAT_GLOBAL,
109 	O2HB_HEARTBEAT_NUM_MODES,
110 };
111 
112 static const char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = {
113 	"local",	/* O2HB_HEARTBEAT_LOCAL */
114 	"global",	/* O2HB_HEARTBEAT_GLOBAL */
115 };
116 
117 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
118 static unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL;
119 
120 /*
121  * o2hb_dependent_users tracks the number of registered callbacks that depend
122  * on heartbeat. o2net and o2dlm are two entities that register this callback.
123  * However only o2dlm depends on the heartbeat. It does not want the heartbeat
124  * to stop while a dlm domain is still active.
125  */
126 static unsigned int o2hb_dependent_users;
127 
128 /*
129  * In global heartbeat mode, all regions are pinned if there are one or more
130  * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All
131  * regions are unpinned if the region count exceeds the cut off or the number
132  * of dependent users falls to zero.
133  */
134 #define O2HB_PIN_CUT_OFF		3
135 
136 /*
137  * In local heartbeat mode, we assume the dlm domain name to be the same as
138  * region uuid. This is true for domains created for the file system but not
139  * necessarily true for userdlm domains. This is a known limitation.
140  *
141  * In global heartbeat mode, we pin/unpin all o2hb regions. This solution
142  * works for both file system and userdlm domains.
143  */
144 static int o2hb_region_pin(const char *region_uuid);
145 static void o2hb_region_unpin(const char *region_uuid);
146 
147 /* Only sets a new threshold if there are no active regions.
148  *
149  * No locking or otherwise interesting code is required for reading
150  * o2hb_dead_threshold as it can't change once regions are active and
151  * it's not interesting to anyone until then anyway. */
152 static void o2hb_dead_threshold_set(unsigned int threshold)
153 {
154 	if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
155 		spin_lock(&o2hb_live_lock);
156 		if (list_empty(&o2hb_all_regions))
157 			o2hb_dead_threshold = threshold;
158 		spin_unlock(&o2hb_live_lock);
159 	}
160 }
161 
162 static int o2hb_global_heartbeat_mode_set(unsigned int hb_mode)
163 {
164 	int ret = -1;
165 
166 	if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) {
167 		spin_lock(&o2hb_live_lock);
168 		if (list_empty(&o2hb_all_regions)) {
169 			o2hb_heartbeat_mode = hb_mode;
170 			ret = 0;
171 		}
172 		spin_unlock(&o2hb_live_lock);
173 	}
174 
175 	return ret;
176 }
177 
178 struct o2hb_node_event {
179 	struct list_head        hn_item;
180 	enum o2hb_callback_type hn_event_type;
181 	struct o2nm_node        *hn_node;
182 	int                     hn_node_num;
183 };
184 
185 struct o2hb_disk_slot {
186 	struct o2hb_disk_heartbeat_block *ds_raw_block;
187 	u8			ds_node_num;
188 	u64			ds_last_time;
189 	u64			ds_last_generation;
190 	u16			ds_equal_samples;
191 	u16			ds_changed_samples;
192 	struct list_head	ds_live_item;
193 };
194 
195 /* each thread owns a region.. when we're asked to tear down the region
196  * we ask the thread to stop, who cleans up the region */
197 struct o2hb_region {
198 	struct config_item	hr_item;
199 
200 	struct list_head	hr_all_item;
201 	unsigned		hr_unclean_stop:1,
202 				hr_aborted_start:1,
203 				hr_item_pinned:1,
204 				hr_item_dropped:1,
205 				hr_node_deleted:1;
206 
207 	/* protected by the hr_callback_sem */
208 	struct task_struct 	*hr_task;
209 
210 	unsigned int		hr_blocks;
211 	unsigned long long	hr_start_block;
212 
213 	unsigned int		hr_block_bits;
214 	unsigned int		hr_block_bytes;
215 
216 	unsigned int		hr_slots_per_page;
217 	unsigned int		hr_num_pages;
218 
219 	struct page             **hr_slot_data;
220 	struct block_device	*hr_bdev;
221 	struct o2hb_disk_slot	*hr_slots;
222 
223 	/* live node map of this region */
224 	unsigned long		hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
225 	unsigned int		hr_region_num;
226 
227 	struct dentry		*hr_debug_dir;
228 	struct o2hb_debug_buf	*hr_db_livenodes;
229 	struct o2hb_debug_buf	*hr_db_regnum;
230 	struct o2hb_debug_buf	*hr_db_elapsed_time;
231 	struct o2hb_debug_buf	*hr_db_pinned;
232 
233 	/* let the person setting up hb wait for it to return until it
234 	 * has reached a 'steady' state.  This will be fixed when we have
235 	 * a more complete api that doesn't lead to this sort of fragility. */
236 	atomic_t		hr_steady_iterations;
237 
238 	/* terminate o2hb thread if it does not reach steady state
239 	 * (hr_steady_iterations == 0) within hr_unsteady_iterations */
240 	atomic_t		hr_unsteady_iterations;
241 
242 	char			hr_dev_name[BDEVNAME_SIZE];
243 
244 	unsigned int		hr_timeout_ms;
245 
246 	/* randomized as the region goes up and down so that a node
247 	 * recognizes a node going up and down in one iteration */
248 	u64			hr_generation;
249 
250 	struct delayed_work	hr_write_timeout_work;
251 	unsigned long		hr_last_timeout_start;
252 
253 	/* negotiate timer, used to negotiate extending hb timeout. */
254 	struct delayed_work	hr_nego_timeout_work;
255 	unsigned long		hr_nego_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
256 
257 	/* Used during o2hb_check_slot to hold a copy of the block
258 	 * being checked because we temporarily have to zero out the
259 	 * crc field. */
260 	struct o2hb_disk_heartbeat_block *hr_tmp_block;
261 
262 	/* Message key for negotiate timeout message. */
263 	unsigned int		hr_key;
264 	struct list_head	hr_handler_list;
265 
266 	/* last hb status, 0 for success, other value for error. */
267 	int			hr_last_hb_status;
268 };
269 
270 struct o2hb_bio_wait_ctxt {
271 	atomic_t          wc_num_reqs;
272 	struct completion wc_io_complete;
273 	int               wc_error;
274 };
275 
276 #define O2HB_NEGO_TIMEOUT_MS (O2HB_MAX_WRITE_TIMEOUT_MS/2)
277 
278 enum {
279 	O2HB_NEGO_TIMEOUT_MSG = 1,
280 	O2HB_NEGO_APPROVE_MSG = 2,
281 };
282 
283 struct o2hb_nego_msg {
284 	u8 node_num;
285 };
286 
287 static void o2hb_write_timeout(struct work_struct *work)
288 {
289 	int failed, quorum;
290 	struct o2hb_region *reg =
291 		container_of(work, struct o2hb_region,
292 			     hr_write_timeout_work.work);
293 
294 	mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
295 	     "milliseconds\n", reg->hr_dev_name,
296 	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
297 
298 	if (o2hb_global_heartbeat_active()) {
299 		spin_lock(&o2hb_live_lock);
300 		if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
301 			set_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
302 		failed = bitmap_weight(o2hb_failed_region_bitmap,
303 					O2NM_MAX_REGIONS);
304 		quorum = bitmap_weight(o2hb_quorum_region_bitmap,
305 					O2NM_MAX_REGIONS);
306 		spin_unlock(&o2hb_live_lock);
307 
308 		mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n",
309 		     quorum, failed);
310 
311 		/*
312 		 * Fence if the number of failed regions >= half the number
313 		 * of  quorum regions
314 		 */
315 		if ((failed << 1) < quorum)
316 			return;
317 	}
318 
319 	o2quo_disk_timeout();
320 }
321 
322 static void o2hb_arm_timeout(struct o2hb_region *reg)
323 {
324 	/* Arm writeout only after thread reaches steady state */
325 	if (atomic_read(&reg->hr_steady_iterations) != 0)
326 		return;
327 
328 	mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n",
329 	     O2HB_MAX_WRITE_TIMEOUT_MS);
330 
331 	if (o2hb_global_heartbeat_active()) {
332 		spin_lock(&o2hb_live_lock);
333 		clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
334 		spin_unlock(&o2hb_live_lock);
335 	}
336 	cancel_delayed_work(&reg->hr_write_timeout_work);
337 	schedule_delayed_work(&reg->hr_write_timeout_work,
338 			      msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
339 
340 	cancel_delayed_work(&reg->hr_nego_timeout_work);
341 	/* negotiate timeout must be less than write timeout. */
342 	schedule_delayed_work(&reg->hr_nego_timeout_work,
343 			      msecs_to_jiffies(O2HB_NEGO_TIMEOUT_MS));
344 	memset(reg->hr_nego_node_bitmap, 0, sizeof(reg->hr_nego_node_bitmap));
345 }
346 
347 static void o2hb_disarm_timeout(struct o2hb_region *reg)
348 {
349 	cancel_delayed_work_sync(&reg->hr_write_timeout_work);
350 	cancel_delayed_work_sync(&reg->hr_nego_timeout_work);
351 }
352 
353 static int o2hb_send_nego_msg(int key, int type, u8 target)
354 {
355 	struct o2hb_nego_msg msg;
356 	int status, ret;
357 
358 	msg.node_num = o2nm_this_node();
359 again:
360 	ret = o2net_send_message(type, key, &msg, sizeof(msg),
361 			target, &status);
362 
363 	if (ret == -EAGAIN || ret == -ENOMEM) {
364 		msleep(100);
365 		goto again;
366 	}
367 
368 	return ret;
369 }
370 
371 static void o2hb_nego_timeout(struct work_struct *work)
372 {
373 	unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
374 	int master_node, i, ret;
375 	struct o2hb_region *reg;
376 
377 	reg = container_of(work, struct o2hb_region, hr_nego_timeout_work.work);
378 	/* don't negotiate timeout if last hb failed since it is very
379 	 * possible io failed. Should let write timeout fence self.
380 	 */
381 	if (reg->hr_last_hb_status)
382 		return;
383 
384 	o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
385 	/* lowest node as master node to make negotiate decision. */
386 	master_node = find_next_bit(live_node_bitmap, O2NM_MAX_NODES, 0);
387 
388 	if (master_node == o2nm_this_node()) {
389 		if (!test_bit(master_node, reg->hr_nego_node_bitmap)) {
390 			printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s).\n",
391 				o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000,
392 				config_item_name(&reg->hr_item), reg->hr_dev_name);
393 			set_bit(master_node, reg->hr_nego_node_bitmap);
394 		}
395 		if (memcmp(reg->hr_nego_node_bitmap, live_node_bitmap,
396 				sizeof(reg->hr_nego_node_bitmap))) {
397 			/* check negotiate bitmap every second to do timeout
398 			 * approve decision.
399 			 */
400 			schedule_delayed_work(&reg->hr_nego_timeout_work,
401 				msecs_to_jiffies(1000));
402 
403 			return;
404 		}
405 
406 		printk(KERN_NOTICE "o2hb: all nodes hb write hung, maybe region %s (%s) is down.\n",
407 			config_item_name(&reg->hr_item), reg->hr_dev_name);
408 		/* approve negotiate timeout request. */
409 		o2hb_arm_timeout(reg);
410 
411 		i = -1;
412 		while ((i = find_next_bit(live_node_bitmap,
413 				O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
414 			if (i == master_node)
415 				continue;
416 
417 			mlog(ML_HEARTBEAT, "send NEGO_APPROVE msg to node %d\n", i);
418 			ret = o2hb_send_nego_msg(reg->hr_key,
419 					O2HB_NEGO_APPROVE_MSG, i);
420 			if (ret)
421 				mlog(ML_ERROR, "send NEGO_APPROVE msg to node %d fail %d\n",
422 					i, ret);
423 		}
424 	} else {
425 		/* negotiate timeout with master node. */
426 		printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s), negotiate timeout with node %d.\n",
427 			o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000, config_item_name(&reg->hr_item),
428 			reg->hr_dev_name, master_node);
429 		ret = o2hb_send_nego_msg(reg->hr_key, O2HB_NEGO_TIMEOUT_MSG,
430 				master_node);
431 		if (ret)
432 			mlog(ML_ERROR, "send NEGO_TIMEOUT msg to node %d fail %d\n",
433 				master_node, ret);
434 	}
435 }
436 
437 static int o2hb_nego_timeout_handler(struct o2net_msg *msg, u32 len, void *data,
438 				void **ret_data)
439 {
440 	struct o2hb_region *reg = data;
441 	struct o2hb_nego_msg *nego_msg;
442 
443 	nego_msg = (struct o2hb_nego_msg *)msg->buf;
444 	printk(KERN_NOTICE "o2hb: receive negotiate timeout message from node %d on region %s (%s).\n",
445 		nego_msg->node_num, config_item_name(&reg->hr_item), reg->hr_dev_name);
446 	if (nego_msg->node_num < O2NM_MAX_NODES)
447 		set_bit(nego_msg->node_num, reg->hr_nego_node_bitmap);
448 	else
449 		mlog(ML_ERROR, "got nego timeout message from bad node.\n");
450 
451 	return 0;
452 }
453 
454 static int o2hb_nego_approve_handler(struct o2net_msg *msg, u32 len, void *data,
455 				void **ret_data)
456 {
457 	struct o2hb_region *reg = data;
458 
459 	printk(KERN_NOTICE "o2hb: negotiate timeout approved by master node on region %s (%s).\n",
460 		config_item_name(&reg->hr_item), reg->hr_dev_name);
461 	o2hb_arm_timeout(reg);
462 	return 0;
463 }
464 
465 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc)
466 {
467 	atomic_set(&wc->wc_num_reqs, 1);
468 	init_completion(&wc->wc_io_complete);
469 	wc->wc_error = 0;
470 }
471 
472 /* Used in error paths too */
473 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
474 				     unsigned int num)
475 {
476 	/* sadly atomic_sub_and_test() isn't available on all platforms.  The
477 	 * good news is that the fast path only completes one at a time */
478 	while(num--) {
479 		if (atomic_dec_and_test(&wc->wc_num_reqs)) {
480 			BUG_ON(num > 0);
481 			complete(&wc->wc_io_complete);
482 		}
483 	}
484 }
485 
486 static void o2hb_wait_on_io(struct o2hb_bio_wait_ctxt *wc)
487 {
488 	o2hb_bio_wait_dec(wc, 1);
489 	wait_for_completion(&wc->wc_io_complete);
490 }
491 
492 static void o2hb_bio_end_io(struct bio *bio)
493 {
494 	struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
495 
496 	if (bio->bi_status) {
497 		mlog(ML_ERROR, "IO Error %d\n", bio->bi_status);
498 		wc->wc_error = blk_status_to_errno(bio->bi_status);
499 	}
500 
501 	o2hb_bio_wait_dec(wc, 1);
502 	bio_put(bio);
503 }
504 
505 /* Setup a Bio to cover I/O against num_slots slots starting at
506  * start_slot. */
507 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
508 				      struct o2hb_bio_wait_ctxt *wc,
509 				      unsigned int *current_slot,
510 				      unsigned int max_slots, int op,
511 				      int op_flags)
512 {
513 	int len, current_page;
514 	unsigned int vec_len, vec_start;
515 	unsigned int bits = reg->hr_block_bits;
516 	unsigned int spp = reg->hr_slots_per_page;
517 	unsigned int cs = *current_slot;
518 	struct bio *bio;
519 	struct page *page;
520 
521 	/* Testing has shown this allocation to take long enough under
522 	 * GFP_KERNEL that the local node can get fenced. It would be
523 	 * nicest if we could pre-allocate these bios and avoid this
524 	 * all together. */
525 	bio = bio_alloc(GFP_ATOMIC, 16);
526 	if (!bio) {
527 		mlog(ML_ERROR, "Could not alloc slots BIO!\n");
528 		bio = ERR_PTR(-ENOMEM);
529 		goto bail;
530 	}
531 
532 	/* Must put everything in 512 byte sectors for the bio... */
533 	bio->bi_iter.bi_sector = (reg->hr_start_block + cs) << (bits - 9);
534 	bio_set_dev(bio, reg->hr_bdev);
535 	bio->bi_private = wc;
536 	bio->bi_end_io = o2hb_bio_end_io;
537 	bio_set_op_attrs(bio, op, op_flags);
538 
539 	vec_start = (cs << bits) % PAGE_SIZE;
540 	while(cs < max_slots) {
541 		current_page = cs / spp;
542 		page = reg->hr_slot_data[current_page];
543 
544 		vec_len = min(PAGE_SIZE - vec_start,
545 			      (max_slots-cs) * (PAGE_SIZE/spp) );
546 
547 		mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
548 		     current_page, vec_len, vec_start);
549 
550 		len = bio_add_page(bio, page, vec_len, vec_start);
551 		if (len != vec_len) break;
552 
553 		cs += vec_len / (PAGE_SIZE/spp);
554 		vec_start = 0;
555 	}
556 
557 bail:
558 	*current_slot = cs;
559 	return bio;
560 }
561 
562 static int o2hb_read_slots(struct o2hb_region *reg,
563 			   unsigned int begin_slot,
564 			   unsigned int max_slots)
565 {
566 	unsigned int current_slot = begin_slot;
567 	int status;
568 	struct o2hb_bio_wait_ctxt wc;
569 	struct bio *bio;
570 
571 	o2hb_bio_wait_init(&wc);
572 
573 	while(current_slot < max_slots) {
574 		bio = o2hb_setup_one_bio(reg, &wc, &current_slot, max_slots,
575 					 REQ_OP_READ, 0);
576 		if (IS_ERR(bio)) {
577 			status = PTR_ERR(bio);
578 			mlog_errno(status);
579 			goto bail_and_wait;
580 		}
581 
582 		atomic_inc(&wc.wc_num_reqs);
583 		submit_bio(bio);
584 	}
585 
586 	status = 0;
587 
588 bail_and_wait:
589 	o2hb_wait_on_io(&wc);
590 	if (wc.wc_error && !status)
591 		status = wc.wc_error;
592 
593 	return status;
594 }
595 
596 static int o2hb_issue_node_write(struct o2hb_region *reg,
597 				 struct o2hb_bio_wait_ctxt *write_wc)
598 {
599 	int status;
600 	unsigned int slot;
601 	struct bio *bio;
602 
603 	o2hb_bio_wait_init(write_wc);
604 
605 	slot = o2nm_this_node();
606 
607 	bio = o2hb_setup_one_bio(reg, write_wc, &slot, slot+1, REQ_OP_WRITE,
608 				 REQ_SYNC);
609 	if (IS_ERR(bio)) {
610 		status = PTR_ERR(bio);
611 		mlog_errno(status);
612 		goto bail;
613 	}
614 
615 	atomic_inc(&write_wc->wc_num_reqs);
616 	submit_bio(bio);
617 
618 	status = 0;
619 bail:
620 	return status;
621 }
622 
623 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
624 				     struct o2hb_disk_heartbeat_block *hb_block)
625 {
626 	__le32 old_cksum;
627 	u32 ret;
628 
629 	/* We want to compute the block crc with a 0 value in the
630 	 * hb_cksum field. Save it off here and replace after the
631 	 * crc. */
632 	old_cksum = hb_block->hb_cksum;
633 	hb_block->hb_cksum = 0;
634 
635 	ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
636 
637 	hb_block->hb_cksum = old_cksum;
638 
639 	return ret;
640 }
641 
642 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
643 {
644 	mlog(ML_ERROR, "Dump slot information: seq = 0x%llx, node = %u, "
645 	     "cksum = 0x%x, generation 0x%llx\n",
646 	     (long long)le64_to_cpu(hb_block->hb_seq),
647 	     hb_block->hb_node, le32_to_cpu(hb_block->hb_cksum),
648 	     (long long)le64_to_cpu(hb_block->hb_generation));
649 }
650 
651 static int o2hb_verify_crc(struct o2hb_region *reg,
652 			   struct o2hb_disk_heartbeat_block *hb_block)
653 {
654 	u32 read, computed;
655 
656 	read = le32_to_cpu(hb_block->hb_cksum);
657 	computed = o2hb_compute_block_crc_le(reg, hb_block);
658 
659 	return read == computed;
660 }
661 
662 /*
663  * Compare the slot data with what we wrote in the last iteration.
664  * If the match fails, print an appropriate error message. This is to
665  * detect errors like... another node hearting on the same slot,
666  * flaky device that is losing writes, etc.
667  * Returns 1 if check succeeds, 0 otherwise.
668  */
669 static int o2hb_check_own_slot(struct o2hb_region *reg)
670 {
671 	struct o2hb_disk_slot *slot;
672 	struct o2hb_disk_heartbeat_block *hb_block;
673 	char *errstr;
674 
675 	slot = &reg->hr_slots[o2nm_this_node()];
676 	/* Don't check on our 1st timestamp */
677 	if (!slot->ds_last_time)
678 		return 0;
679 
680 	hb_block = slot->ds_raw_block;
681 	if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time &&
682 	    le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation &&
683 	    hb_block->hb_node == slot->ds_node_num)
684 		return 1;
685 
686 #define ERRSTR1		"Another node is heartbeating on device"
687 #define ERRSTR2		"Heartbeat generation mismatch on device"
688 #define ERRSTR3		"Heartbeat sequence mismatch on device"
689 
690 	if (hb_block->hb_node != slot->ds_node_num)
691 		errstr = ERRSTR1;
692 	else if (le64_to_cpu(hb_block->hb_generation) !=
693 		 slot->ds_last_generation)
694 		errstr = ERRSTR2;
695 	else
696 		errstr = ERRSTR3;
697 
698 	mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), "
699 	     "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name,
700 	     slot->ds_node_num, (unsigned long long)slot->ds_last_generation,
701 	     (unsigned long long)slot->ds_last_time, hb_block->hb_node,
702 	     (unsigned long long)le64_to_cpu(hb_block->hb_generation),
703 	     (unsigned long long)le64_to_cpu(hb_block->hb_seq));
704 
705 	return 0;
706 }
707 
708 static inline void o2hb_prepare_block(struct o2hb_region *reg,
709 				      u64 generation)
710 {
711 	int node_num;
712 	u64 cputime;
713 	struct o2hb_disk_slot *slot;
714 	struct o2hb_disk_heartbeat_block *hb_block;
715 
716 	node_num = o2nm_this_node();
717 	slot = &reg->hr_slots[node_num];
718 
719 	hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
720 	memset(hb_block, 0, reg->hr_block_bytes);
721 	/* TODO: time stuff */
722 	cputime = ktime_get_real_seconds();
723 	if (!cputime)
724 		cputime = 1;
725 
726 	hb_block->hb_seq = cpu_to_le64(cputime);
727 	hb_block->hb_node = node_num;
728 	hb_block->hb_generation = cpu_to_le64(generation);
729 	hb_block->hb_dead_ms = cpu_to_le32(o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS);
730 
731 	/* This step must always happen last! */
732 	hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
733 								   hb_block));
734 
735 	mlog(ML_HB_BIO, "our node generation = 0x%llx, cksum = 0x%x\n",
736 	     (long long)generation,
737 	     le32_to_cpu(hb_block->hb_cksum));
738 }
739 
740 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
741 				struct o2nm_node *node,
742 				int idx)
743 {
744 	struct o2hb_callback_func *f;
745 
746 	list_for_each_entry(f, &hbcall->list, hc_item) {
747 		mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
748 		(f->hc_func)(node, idx, f->hc_data);
749 	}
750 }
751 
752 /* Will run the list in order until we process the passed event */
753 static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
754 {
755 	struct o2hb_callback *hbcall;
756 	struct o2hb_node_event *event;
757 
758 	/* Holding callback sem assures we don't alter the callback
759 	 * lists when doing this, and serializes ourselves with other
760 	 * processes wanting callbacks. */
761 	down_write(&o2hb_callback_sem);
762 
763 	spin_lock(&o2hb_live_lock);
764 	while (!list_empty(&o2hb_node_events)
765 	       && !list_empty(&queued_event->hn_item)) {
766 		event = list_entry(o2hb_node_events.next,
767 				   struct o2hb_node_event,
768 				   hn_item);
769 		list_del_init(&event->hn_item);
770 		spin_unlock(&o2hb_live_lock);
771 
772 		mlog(ML_HEARTBEAT, "Node %s event for %d\n",
773 		     event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
774 		     event->hn_node_num);
775 
776 		hbcall = hbcall_from_type(event->hn_event_type);
777 
778 		/* We should *never* have gotten on to the list with a
779 		 * bad type... This isn't something that we should try
780 		 * to recover from. */
781 		BUG_ON(IS_ERR(hbcall));
782 
783 		o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
784 
785 		spin_lock(&o2hb_live_lock);
786 	}
787 	spin_unlock(&o2hb_live_lock);
788 
789 	up_write(&o2hb_callback_sem);
790 }
791 
792 static void o2hb_queue_node_event(struct o2hb_node_event *event,
793 				  enum o2hb_callback_type type,
794 				  struct o2nm_node *node,
795 				  int node_num)
796 {
797 	assert_spin_locked(&o2hb_live_lock);
798 
799 	BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB));
800 
801 	event->hn_event_type = type;
802 	event->hn_node = node;
803 	event->hn_node_num = node_num;
804 
805 	mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
806 	     type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
807 
808 	list_add_tail(&event->hn_item, &o2hb_node_events);
809 }
810 
811 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
812 {
813 	struct o2hb_node_event event =
814 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
815 	struct o2nm_node *node;
816 	int queued = 0;
817 
818 	node = o2nm_get_node_by_num(slot->ds_node_num);
819 	if (!node)
820 		return;
821 
822 	spin_lock(&o2hb_live_lock);
823 	if (!list_empty(&slot->ds_live_item)) {
824 		mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
825 		     slot->ds_node_num);
826 
827 		list_del_init(&slot->ds_live_item);
828 
829 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
830 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
831 
832 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
833 					      slot->ds_node_num);
834 			queued = 1;
835 		}
836 	}
837 	spin_unlock(&o2hb_live_lock);
838 
839 	if (queued)
840 		o2hb_run_event_list(&event);
841 
842 	o2nm_node_put(node);
843 }
844 
845 static void o2hb_set_quorum_device(struct o2hb_region *reg)
846 {
847 	if (!o2hb_global_heartbeat_active())
848 		return;
849 
850 	/* Prevent race with o2hb_heartbeat_group_drop_item() */
851 	if (kthread_should_stop())
852 		return;
853 
854 	/* Tag region as quorum only after thread reaches steady state */
855 	if (atomic_read(&reg->hr_steady_iterations) != 0)
856 		return;
857 
858 	spin_lock(&o2hb_live_lock);
859 
860 	if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
861 		goto unlock;
862 
863 	/*
864 	 * A region can be added to the quorum only when it sees all
865 	 * live nodes heartbeat on it. In other words, the region has been
866 	 * added to all nodes.
867 	 */
868 	if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap,
869 		   sizeof(o2hb_live_node_bitmap)))
870 		goto unlock;
871 
872 	printk(KERN_NOTICE "o2hb: Region %s (%s) is now a quorum device\n",
873 	       config_item_name(&reg->hr_item), reg->hr_dev_name);
874 
875 	set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
876 
877 	/*
878 	 * If global heartbeat active, unpin all regions if the
879 	 * region count > CUT_OFF
880 	 */
881 	if (bitmap_weight(o2hb_quorum_region_bitmap,
882 			   O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF)
883 		o2hb_region_unpin(NULL);
884 unlock:
885 	spin_unlock(&o2hb_live_lock);
886 }
887 
888 static int o2hb_check_slot(struct o2hb_region *reg,
889 			   struct o2hb_disk_slot *slot)
890 {
891 	int changed = 0, gen_changed = 0;
892 	struct o2hb_node_event event =
893 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
894 	struct o2nm_node *node;
895 	struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
896 	u64 cputime;
897 	unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS;
898 	unsigned int slot_dead_ms;
899 	int tmp;
900 	int queued = 0;
901 
902 	memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
903 
904 	/*
905 	 * If a node is no longer configured but is still in the livemap, we
906 	 * may need to clear that bit from the livemap.
907 	 */
908 	node = o2nm_get_node_by_num(slot->ds_node_num);
909 	if (!node) {
910 		spin_lock(&o2hb_live_lock);
911 		tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap);
912 		spin_unlock(&o2hb_live_lock);
913 		if (!tmp)
914 			return 0;
915 	}
916 
917 	if (!o2hb_verify_crc(reg, hb_block)) {
918 		/* all paths from here will drop o2hb_live_lock for
919 		 * us. */
920 		spin_lock(&o2hb_live_lock);
921 
922 		/* Don't print an error on the console in this case -
923 		 * a freshly formatted heartbeat area will not have a
924 		 * crc set on it. */
925 		if (list_empty(&slot->ds_live_item))
926 			goto out;
927 
928 		/* The node is live but pushed out a bad crc. We
929 		 * consider it a transient miss but don't populate any
930 		 * other values as they may be junk. */
931 		mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
932 		     slot->ds_node_num, reg->hr_dev_name);
933 		o2hb_dump_slot(hb_block);
934 
935 		slot->ds_equal_samples++;
936 		goto fire_callbacks;
937 	}
938 
939 	/* we don't care if these wrap.. the state transitions below
940 	 * clear at the right places */
941 	cputime = le64_to_cpu(hb_block->hb_seq);
942 	if (slot->ds_last_time != cputime)
943 		slot->ds_changed_samples++;
944 	else
945 		slot->ds_equal_samples++;
946 	slot->ds_last_time = cputime;
947 
948 	/* The node changed heartbeat generations. We assume this to
949 	 * mean it dropped off but came back before we timed out. We
950 	 * want to consider it down for the time being but don't want
951 	 * to lose any changed_samples state we might build up to
952 	 * considering it live again. */
953 	if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
954 		gen_changed = 1;
955 		slot->ds_equal_samples = 0;
956 		mlog(ML_HEARTBEAT, "Node %d changed generation (0x%llx "
957 		     "to 0x%llx)\n", slot->ds_node_num,
958 		     (long long)slot->ds_last_generation,
959 		     (long long)le64_to_cpu(hb_block->hb_generation));
960 	}
961 
962 	slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
963 
964 	mlog(ML_HEARTBEAT, "Slot %d gen 0x%llx cksum 0x%x "
965 	     "seq %llu last %llu changed %u equal %u\n",
966 	     slot->ds_node_num, (long long)slot->ds_last_generation,
967 	     le32_to_cpu(hb_block->hb_cksum),
968 	     (unsigned long long)le64_to_cpu(hb_block->hb_seq),
969 	     (unsigned long long)slot->ds_last_time, slot->ds_changed_samples,
970 	     slot->ds_equal_samples);
971 
972 	spin_lock(&o2hb_live_lock);
973 
974 fire_callbacks:
975 	/* dead nodes only come to life after some number of
976 	 * changes at any time during their dead time */
977 	if (list_empty(&slot->ds_live_item) &&
978 	    slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
979 		mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n",
980 		     slot->ds_node_num, (long long)slot->ds_last_generation);
981 
982 		set_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
983 
984 		/* first on the list generates a callback */
985 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
986 			mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes "
987 			     "bitmap\n", slot->ds_node_num);
988 			set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
989 
990 			o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
991 					      slot->ds_node_num);
992 
993 			changed = 1;
994 			queued = 1;
995 		}
996 
997 		list_add_tail(&slot->ds_live_item,
998 			      &o2hb_live_slots[slot->ds_node_num]);
999 
1000 		slot->ds_equal_samples = 0;
1001 
1002 		/* We want to be sure that all nodes agree on the
1003 		 * number of milliseconds before a node will be
1004 		 * considered dead. The self-fencing timeout is
1005 		 * computed from this value, and a discrepancy might
1006 		 * result in heartbeat calling a node dead when it
1007 		 * hasn't self-fenced yet. */
1008 		slot_dead_ms = le32_to_cpu(hb_block->hb_dead_ms);
1009 		if (slot_dead_ms && slot_dead_ms != dead_ms) {
1010 			/* TODO: Perhaps we can fail the region here. */
1011 			mlog(ML_ERROR, "Node %d on device %s has a dead count "
1012 			     "of %u ms, but our count is %u ms.\n"
1013 			     "Please double check your configuration values "
1014 			     "for 'O2CB_HEARTBEAT_THRESHOLD'\n",
1015 			     slot->ds_node_num, reg->hr_dev_name, slot_dead_ms,
1016 			     dead_ms);
1017 		}
1018 		goto out;
1019 	}
1020 
1021 	/* if the list is dead, we're done.. */
1022 	if (list_empty(&slot->ds_live_item))
1023 		goto out;
1024 
1025 	/* live nodes only go dead after enough consequtive missed
1026 	 * samples..  reset the missed counter whenever we see
1027 	 * activity */
1028 	if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
1029 		mlog(ML_HEARTBEAT, "Node %d left my region\n",
1030 		     slot->ds_node_num);
1031 
1032 		clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
1033 
1034 		/* last off the live_slot generates a callback */
1035 		list_del_init(&slot->ds_live_item);
1036 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
1037 			mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live "
1038 			     "nodes bitmap\n", slot->ds_node_num);
1039 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
1040 
1041 			/* node can be null */
1042 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB,
1043 					      node, slot->ds_node_num);
1044 
1045 			changed = 1;
1046 			queued = 1;
1047 		}
1048 
1049 		/* We don't clear this because the node is still
1050 		 * actually writing new blocks. */
1051 		if (!gen_changed)
1052 			slot->ds_changed_samples = 0;
1053 		goto out;
1054 	}
1055 	if (slot->ds_changed_samples) {
1056 		slot->ds_changed_samples = 0;
1057 		slot->ds_equal_samples = 0;
1058 	}
1059 out:
1060 	spin_unlock(&o2hb_live_lock);
1061 
1062 	if (queued)
1063 		o2hb_run_event_list(&event);
1064 
1065 	if (node)
1066 		o2nm_node_put(node);
1067 	return changed;
1068 }
1069 
1070 static int o2hb_highest_node(unsigned long *nodes, int numbits)
1071 {
1072 	return find_last_bit(nodes, numbits);
1073 }
1074 
1075 static int o2hb_lowest_node(unsigned long *nodes, int numbits)
1076 {
1077 	return find_first_bit(nodes, numbits);
1078 }
1079 
1080 static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
1081 {
1082 	int i, ret, highest_node, lowest_node;
1083 	int membership_change = 0, own_slot_ok = 0;
1084 	unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
1085 	unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
1086 	struct o2hb_bio_wait_ctxt write_wc;
1087 
1088 	ret = o2nm_configured_node_map(configured_nodes,
1089 				       sizeof(configured_nodes));
1090 	if (ret) {
1091 		mlog_errno(ret);
1092 		goto bail;
1093 	}
1094 
1095 	/*
1096 	 * If a node is not configured but is in the livemap, we still need
1097 	 * to read the slot so as to be able to remove it from the livemap.
1098 	 */
1099 	o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
1100 	i = -1;
1101 	while ((i = find_next_bit(live_node_bitmap,
1102 				  O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1103 		set_bit(i, configured_nodes);
1104 	}
1105 
1106 	highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
1107 	lowest_node = o2hb_lowest_node(configured_nodes, O2NM_MAX_NODES);
1108 	if (highest_node >= O2NM_MAX_NODES || lowest_node >= O2NM_MAX_NODES) {
1109 		mlog(ML_NOTICE, "o2hb: No configured nodes found!\n");
1110 		ret = -EINVAL;
1111 		goto bail;
1112 	}
1113 
1114 	/* No sense in reading the slots of nodes that don't exist
1115 	 * yet. Of course, if the node definitions have holes in them
1116 	 * then we're reading an empty slot anyway... Consider this
1117 	 * best-effort. */
1118 	ret = o2hb_read_slots(reg, lowest_node, highest_node + 1);
1119 	if (ret < 0) {
1120 		mlog_errno(ret);
1121 		goto bail;
1122 	}
1123 
1124 	/* With an up to date view of the slots, we can check that no
1125 	 * other node has been improperly configured to heartbeat in
1126 	 * our slot. */
1127 	own_slot_ok = o2hb_check_own_slot(reg);
1128 
1129 	/* fill in the proper info for our next heartbeat */
1130 	o2hb_prepare_block(reg, reg->hr_generation);
1131 
1132 	ret = o2hb_issue_node_write(reg, &write_wc);
1133 	if (ret < 0) {
1134 		mlog_errno(ret);
1135 		goto bail;
1136 	}
1137 
1138 	i = -1;
1139 	while((i = find_next_bit(configured_nodes,
1140 				 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1141 		membership_change |= o2hb_check_slot(reg, &reg->hr_slots[i]);
1142 	}
1143 
1144 	/*
1145 	 * We have to be sure we've advertised ourselves on disk
1146 	 * before we can go to steady state.  This ensures that
1147 	 * people we find in our steady state have seen us.
1148 	 */
1149 	o2hb_wait_on_io(&write_wc);
1150 	if (write_wc.wc_error) {
1151 		/* Do not re-arm the write timeout on I/O error - we
1152 		 * can't be sure that the new block ever made it to
1153 		 * disk */
1154 		mlog(ML_ERROR, "Write error %d on device \"%s\"\n",
1155 		     write_wc.wc_error, reg->hr_dev_name);
1156 		ret = write_wc.wc_error;
1157 		goto bail;
1158 	}
1159 
1160 	/* Skip disarming the timeout if own slot has stale/bad data */
1161 	if (own_slot_ok) {
1162 		o2hb_set_quorum_device(reg);
1163 		o2hb_arm_timeout(reg);
1164 		reg->hr_last_timeout_start = jiffies;
1165 	}
1166 
1167 bail:
1168 	/* let the person who launched us know when things are steady */
1169 	if (atomic_read(&reg->hr_steady_iterations) != 0) {
1170 		if (!ret && own_slot_ok && !membership_change) {
1171 			if (atomic_dec_and_test(&reg->hr_steady_iterations))
1172 				wake_up(&o2hb_steady_queue);
1173 		}
1174 	}
1175 
1176 	if (atomic_read(&reg->hr_steady_iterations) != 0) {
1177 		if (atomic_dec_and_test(&reg->hr_unsteady_iterations)) {
1178 			printk(KERN_NOTICE "o2hb: Unable to stabilize "
1179 			       "heartbeat on region %s (%s)\n",
1180 			       config_item_name(&reg->hr_item),
1181 			       reg->hr_dev_name);
1182 			atomic_set(&reg->hr_steady_iterations, 0);
1183 			reg->hr_aborted_start = 1;
1184 			wake_up(&o2hb_steady_queue);
1185 			ret = -EIO;
1186 		}
1187 	}
1188 
1189 	return ret;
1190 }
1191 
1192 /*
1193  * we ride the region ref that the region dir holds.  before the region
1194  * dir is removed and drops it ref it will wait to tear down this
1195  * thread.
1196  */
1197 static int o2hb_thread(void *data)
1198 {
1199 	int i, ret;
1200 	struct o2hb_region *reg = data;
1201 	struct o2hb_bio_wait_ctxt write_wc;
1202 	ktime_t before_hb, after_hb;
1203 	unsigned int elapsed_msec;
1204 
1205 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
1206 
1207 	set_user_nice(current, MIN_NICE);
1208 
1209 	/* Pin node */
1210 	ret = o2nm_depend_this_node();
1211 	if (ret) {
1212 		mlog(ML_ERROR, "Node has been deleted, ret = %d\n", ret);
1213 		reg->hr_node_deleted = 1;
1214 		wake_up(&o2hb_steady_queue);
1215 		return 0;
1216 	}
1217 
1218 	while (!kthread_should_stop() &&
1219 	       !reg->hr_unclean_stop && !reg->hr_aborted_start) {
1220 		/* We track the time spent inside
1221 		 * o2hb_do_disk_heartbeat so that we avoid more than
1222 		 * hr_timeout_ms between disk writes. On busy systems
1223 		 * this should result in a heartbeat which is less
1224 		 * likely to time itself out. */
1225 		before_hb = ktime_get_real();
1226 
1227 		ret = o2hb_do_disk_heartbeat(reg);
1228 		reg->hr_last_hb_status = ret;
1229 
1230 		after_hb = ktime_get_real();
1231 
1232 		elapsed_msec = (unsigned int)
1233 				ktime_ms_delta(after_hb, before_hb);
1234 
1235 		mlog(ML_HEARTBEAT,
1236 		     "start = %lld, end = %lld, msec = %u, ret = %d\n",
1237 		     before_hb, after_hb, elapsed_msec, ret);
1238 
1239 		if (!kthread_should_stop() &&
1240 		    elapsed_msec < reg->hr_timeout_ms) {
1241 			/* the kthread api has blocked signals for us so no
1242 			 * need to record the return value. */
1243 			msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
1244 		}
1245 	}
1246 
1247 	o2hb_disarm_timeout(reg);
1248 
1249 	/* unclean stop is only used in very bad situation */
1250 	for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
1251 		o2hb_shutdown_slot(&reg->hr_slots[i]);
1252 
1253 	/* Explicit down notification - avoid forcing the other nodes
1254 	 * to timeout on this region when we could just as easily
1255 	 * write a clear generation - thus indicating to them that
1256 	 * this node has left this region.
1257 	 */
1258 	if (!reg->hr_unclean_stop && !reg->hr_aborted_start) {
1259 		o2hb_prepare_block(reg, 0);
1260 		ret = o2hb_issue_node_write(reg, &write_wc);
1261 		if (ret == 0)
1262 			o2hb_wait_on_io(&write_wc);
1263 		else
1264 			mlog_errno(ret);
1265 	}
1266 
1267 	/* Unpin node */
1268 	o2nm_undepend_this_node();
1269 
1270 	mlog(ML_HEARTBEAT|ML_KTHREAD, "o2hb thread exiting\n");
1271 
1272 	return 0;
1273 }
1274 
1275 #ifdef CONFIG_DEBUG_FS
1276 static int o2hb_debug_open(struct inode *inode, struct file *file)
1277 {
1278 	struct o2hb_debug_buf *db = inode->i_private;
1279 	struct o2hb_region *reg;
1280 	unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1281 	unsigned long lts;
1282 	char *buf = NULL;
1283 	int i = -1;
1284 	int out = 0;
1285 
1286 	/* max_nodes should be the largest bitmap we pass here */
1287 	BUG_ON(sizeof(map) < db->db_size);
1288 
1289 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
1290 	if (!buf)
1291 		goto bail;
1292 
1293 	switch (db->db_type) {
1294 	case O2HB_DB_TYPE_LIVENODES:
1295 	case O2HB_DB_TYPE_LIVEREGIONS:
1296 	case O2HB_DB_TYPE_QUORUMREGIONS:
1297 	case O2HB_DB_TYPE_FAILEDREGIONS:
1298 		spin_lock(&o2hb_live_lock);
1299 		memcpy(map, db->db_data, db->db_size);
1300 		spin_unlock(&o2hb_live_lock);
1301 		break;
1302 
1303 	case O2HB_DB_TYPE_REGION_LIVENODES:
1304 		spin_lock(&o2hb_live_lock);
1305 		reg = (struct o2hb_region *)db->db_data;
1306 		memcpy(map, reg->hr_live_node_bitmap, db->db_size);
1307 		spin_unlock(&o2hb_live_lock);
1308 		break;
1309 
1310 	case O2HB_DB_TYPE_REGION_NUMBER:
1311 		reg = (struct o2hb_region *)db->db_data;
1312 		out += snprintf(buf + out, PAGE_SIZE - out, "%d\n",
1313 				reg->hr_region_num);
1314 		goto done;
1315 
1316 	case O2HB_DB_TYPE_REGION_ELAPSED_TIME:
1317 		reg = (struct o2hb_region *)db->db_data;
1318 		lts = reg->hr_last_timeout_start;
1319 		/* If 0, it has never been set before */
1320 		if (lts)
1321 			lts = jiffies_to_msecs(jiffies - lts);
1322 		out += snprintf(buf + out, PAGE_SIZE - out, "%lu\n", lts);
1323 		goto done;
1324 
1325 	case O2HB_DB_TYPE_REGION_PINNED:
1326 		reg = (struct o2hb_region *)db->db_data;
1327 		out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
1328 				!!reg->hr_item_pinned);
1329 		goto done;
1330 
1331 	default:
1332 		goto done;
1333 	}
1334 
1335 	while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len)
1336 		out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i);
1337 	out += snprintf(buf + out, PAGE_SIZE - out, "\n");
1338 
1339 done:
1340 	i_size_write(inode, out);
1341 
1342 	file->private_data = buf;
1343 
1344 	return 0;
1345 bail:
1346 	return -ENOMEM;
1347 }
1348 
1349 static int o2hb_debug_release(struct inode *inode, struct file *file)
1350 {
1351 	kfree(file->private_data);
1352 	return 0;
1353 }
1354 
1355 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1356 				 size_t nbytes, loff_t *ppos)
1357 {
1358 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
1359 				       i_size_read(file->f_mapping->host));
1360 }
1361 #else
1362 static int o2hb_debug_open(struct inode *inode, struct file *file)
1363 {
1364 	return 0;
1365 }
1366 static int o2hb_debug_release(struct inode *inode, struct file *file)
1367 {
1368 	return 0;
1369 }
1370 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1371 			       size_t nbytes, loff_t *ppos)
1372 {
1373 	return 0;
1374 }
1375 #endif  /* CONFIG_DEBUG_FS */
1376 
1377 static const struct file_operations o2hb_debug_fops = {
1378 	.open =		o2hb_debug_open,
1379 	.release =	o2hb_debug_release,
1380 	.read =		o2hb_debug_read,
1381 	.llseek =	generic_file_llseek,
1382 };
1383 
1384 void o2hb_exit(void)
1385 {
1386 	debugfs_remove_recursive(o2hb_debug_dir);
1387 	kfree(o2hb_db_livenodes);
1388 	kfree(o2hb_db_liveregions);
1389 	kfree(o2hb_db_quorumregions);
1390 	kfree(o2hb_db_failedregions);
1391 }
1392 
1393 static void o2hb_debug_create(const char *name, struct dentry *dir,
1394 			      struct o2hb_debug_buf **db, int db_len, int type,
1395 			      int size, int len, void *data)
1396 {
1397 	*db = kmalloc(db_len, GFP_KERNEL);
1398 	if (!*db)
1399 		return;
1400 
1401 	(*db)->db_type = type;
1402 	(*db)->db_size = size;
1403 	(*db)->db_len = len;
1404 	(*db)->db_data = data;
1405 
1406 	debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db, &o2hb_debug_fops);
1407 }
1408 
1409 static void o2hb_debug_init(void)
1410 {
1411 	o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL);
1412 
1413 	o2hb_debug_create(O2HB_DEBUG_LIVENODES, o2hb_debug_dir,
1414 			  &o2hb_db_livenodes, sizeof(*o2hb_db_livenodes),
1415 			  O2HB_DB_TYPE_LIVENODES, sizeof(o2hb_live_node_bitmap),
1416 			  O2NM_MAX_NODES, o2hb_live_node_bitmap);
1417 
1418 	o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS, o2hb_debug_dir,
1419 			  &o2hb_db_liveregions, sizeof(*o2hb_db_liveregions),
1420 			  O2HB_DB_TYPE_LIVEREGIONS,
1421 			  sizeof(o2hb_live_region_bitmap), O2NM_MAX_REGIONS,
1422 			  o2hb_live_region_bitmap);
1423 
1424 	o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS, o2hb_debug_dir,
1425 			  &o2hb_db_quorumregions,
1426 			  sizeof(*o2hb_db_quorumregions),
1427 			  O2HB_DB_TYPE_QUORUMREGIONS,
1428 			  sizeof(o2hb_quorum_region_bitmap), O2NM_MAX_REGIONS,
1429 			  o2hb_quorum_region_bitmap);
1430 
1431 	o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS, o2hb_debug_dir,
1432 			  &o2hb_db_failedregions,
1433 			  sizeof(*o2hb_db_failedregions),
1434 			  O2HB_DB_TYPE_FAILEDREGIONS,
1435 			  sizeof(o2hb_failed_region_bitmap), O2NM_MAX_REGIONS,
1436 			  o2hb_failed_region_bitmap);
1437 }
1438 
1439 void o2hb_init(void)
1440 {
1441 	int i;
1442 
1443 	for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
1444 		INIT_LIST_HEAD(&o2hb_callbacks[i].list);
1445 
1446 	for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
1447 		INIT_LIST_HEAD(&o2hb_live_slots[i]);
1448 
1449 	INIT_LIST_HEAD(&o2hb_node_events);
1450 
1451 	memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
1452 	memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap));
1453 	memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap));
1454 	memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap));
1455 	memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap));
1456 
1457 	o2hb_dependent_users = 0;
1458 
1459 	o2hb_debug_init();
1460 }
1461 
1462 /* if we're already in a callback then we're already serialized by the sem */
1463 static void o2hb_fill_node_map_from_callback(unsigned long *map,
1464 					     unsigned bytes)
1465 {
1466 	BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
1467 
1468 	memcpy(map, &o2hb_live_node_bitmap, bytes);
1469 }
1470 
1471 /*
1472  * get a map of all nodes that are heartbeating in any regions
1473  */
1474 void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
1475 {
1476 	/* callers want to serialize this map and callbacks so that they
1477 	 * can trust that they don't miss nodes coming to the party */
1478 	down_read(&o2hb_callback_sem);
1479 	spin_lock(&o2hb_live_lock);
1480 	o2hb_fill_node_map_from_callback(map, bytes);
1481 	spin_unlock(&o2hb_live_lock);
1482 	up_read(&o2hb_callback_sem);
1483 }
1484 EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
1485 
1486 /*
1487  * heartbeat configfs bits.  The heartbeat set is a default set under
1488  * the cluster set in nodemanager.c.
1489  */
1490 
1491 static struct o2hb_region *to_o2hb_region(struct config_item *item)
1492 {
1493 	return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
1494 }
1495 
1496 /* drop_item only drops its ref after killing the thread, nothing should
1497  * be using the region anymore.  this has to clean up any state that
1498  * attributes might have built up. */
1499 static void o2hb_region_release(struct config_item *item)
1500 {
1501 	int i;
1502 	struct page *page;
1503 	struct o2hb_region *reg = to_o2hb_region(item);
1504 
1505 	mlog(ML_HEARTBEAT, "hb region release (%s)\n", reg->hr_dev_name);
1506 
1507 	kfree(reg->hr_tmp_block);
1508 
1509 	if (reg->hr_slot_data) {
1510 		for (i = 0; i < reg->hr_num_pages; i++) {
1511 			page = reg->hr_slot_data[i];
1512 			if (page)
1513 				__free_page(page);
1514 		}
1515 		kfree(reg->hr_slot_data);
1516 	}
1517 
1518 	if (reg->hr_bdev)
1519 		blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1520 
1521 	kfree(reg->hr_slots);
1522 
1523 	debugfs_remove_recursive(reg->hr_debug_dir);
1524 	kfree(reg->hr_db_livenodes);
1525 	kfree(reg->hr_db_regnum);
1526 	kfree(reg->hr_db_elapsed_time);
1527 	kfree(reg->hr_db_pinned);
1528 
1529 	spin_lock(&o2hb_live_lock);
1530 	list_del(&reg->hr_all_item);
1531 	spin_unlock(&o2hb_live_lock);
1532 
1533 	o2net_unregister_handler_list(&reg->hr_handler_list);
1534 	kfree(reg);
1535 }
1536 
1537 static int o2hb_read_block_input(struct o2hb_region *reg,
1538 				 const char *page,
1539 				 unsigned long *ret_bytes,
1540 				 unsigned int *ret_bits)
1541 {
1542 	unsigned long bytes;
1543 	char *p = (char *)page;
1544 
1545 	bytes = simple_strtoul(p, &p, 0);
1546 	if (!p || (*p && (*p != '\n')))
1547 		return -EINVAL;
1548 
1549 	/* Heartbeat and fs min / max block sizes are the same. */
1550 	if (bytes > 4096 || bytes < 512)
1551 		return -ERANGE;
1552 	if (hweight16(bytes) != 1)
1553 		return -EINVAL;
1554 
1555 	if (ret_bytes)
1556 		*ret_bytes = bytes;
1557 	if (ret_bits)
1558 		*ret_bits = ffs(bytes) - 1;
1559 
1560 	return 0;
1561 }
1562 
1563 static ssize_t o2hb_region_block_bytes_show(struct config_item *item,
1564 					    char *page)
1565 {
1566 	return sprintf(page, "%u\n", to_o2hb_region(item)->hr_block_bytes);
1567 }
1568 
1569 static ssize_t o2hb_region_block_bytes_store(struct config_item *item,
1570 					     const char *page,
1571 					     size_t count)
1572 {
1573 	struct o2hb_region *reg = to_o2hb_region(item);
1574 	int status;
1575 	unsigned long block_bytes;
1576 	unsigned int block_bits;
1577 
1578 	if (reg->hr_bdev)
1579 		return -EINVAL;
1580 
1581 	status = o2hb_read_block_input(reg, page, &block_bytes,
1582 				       &block_bits);
1583 	if (status)
1584 		return status;
1585 
1586 	reg->hr_block_bytes = (unsigned int)block_bytes;
1587 	reg->hr_block_bits = block_bits;
1588 
1589 	return count;
1590 }
1591 
1592 static ssize_t o2hb_region_start_block_show(struct config_item *item,
1593 					    char *page)
1594 {
1595 	return sprintf(page, "%llu\n", to_o2hb_region(item)->hr_start_block);
1596 }
1597 
1598 static ssize_t o2hb_region_start_block_store(struct config_item *item,
1599 					     const char *page,
1600 					     size_t count)
1601 {
1602 	struct o2hb_region *reg = to_o2hb_region(item);
1603 	unsigned long long tmp;
1604 	char *p = (char *)page;
1605 
1606 	if (reg->hr_bdev)
1607 		return -EINVAL;
1608 
1609 	tmp = simple_strtoull(p, &p, 0);
1610 	if (!p || (*p && (*p != '\n')))
1611 		return -EINVAL;
1612 
1613 	reg->hr_start_block = tmp;
1614 
1615 	return count;
1616 }
1617 
1618 static ssize_t o2hb_region_blocks_show(struct config_item *item, char *page)
1619 {
1620 	return sprintf(page, "%d\n", to_o2hb_region(item)->hr_blocks);
1621 }
1622 
1623 static ssize_t o2hb_region_blocks_store(struct config_item *item,
1624 					const char *page,
1625 					size_t count)
1626 {
1627 	struct o2hb_region *reg = to_o2hb_region(item);
1628 	unsigned long tmp;
1629 	char *p = (char *)page;
1630 
1631 	if (reg->hr_bdev)
1632 		return -EINVAL;
1633 
1634 	tmp = simple_strtoul(p, &p, 0);
1635 	if (!p || (*p && (*p != '\n')))
1636 		return -EINVAL;
1637 
1638 	if (tmp > O2NM_MAX_NODES || tmp == 0)
1639 		return -ERANGE;
1640 
1641 	reg->hr_blocks = (unsigned int)tmp;
1642 
1643 	return count;
1644 }
1645 
1646 static ssize_t o2hb_region_dev_show(struct config_item *item, char *page)
1647 {
1648 	unsigned int ret = 0;
1649 
1650 	if (to_o2hb_region(item)->hr_bdev)
1651 		ret = sprintf(page, "%s\n", to_o2hb_region(item)->hr_dev_name);
1652 
1653 	return ret;
1654 }
1655 
1656 static void o2hb_init_region_params(struct o2hb_region *reg)
1657 {
1658 	reg->hr_slots_per_page = PAGE_SIZE >> reg->hr_block_bits;
1659 	reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
1660 
1661 	mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
1662 	     reg->hr_start_block, reg->hr_blocks);
1663 	mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
1664 	     reg->hr_block_bytes, reg->hr_block_bits);
1665 	mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
1666 	mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
1667 }
1668 
1669 static int o2hb_map_slot_data(struct o2hb_region *reg)
1670 {
1671 	int i, j;
1672 	unsigned int last_slot;
1673 	unsigned int spp = reg->hr_slots_per_page;
1674 	struct page *page;
1675 	char *raw;
1676 	struct o2hb_disk_slot *slot;
1677 
1678 	reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
1679 	if (reg->hr_tmp_block == NULL)
1680 		return -ENOMEM;
1681 
1682 	reg->hr_slots = kcalloc(reg->hr_blocks,
1683 				sizeof(struct o2hb_disk_slot), GFP_KERNEL);
1684 	if (reg->hr_slots == NULL)
1685 		return -ENOMEM;
1686 
1687 	for(i = 0; i < reg->hr_blocks; i++) {
1688 		slot = &reg->hr_slots[i];
1689 		slot->ds_node_num = i;
1690 		INIT_LIST_HEAD(&slot->ds_live_item);
1691 		slot->ds_raw_block = NULL;
1692 	}
1693 
1694 	reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
1695 	mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
1696 			   "at %u blocks per page\n",
1697 	     reg->hr_num_pages, reg->hr_blocks, spp);
1698 
1699 	reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
1700 				    GFP_KERNEL);
1701 	if (!reg->hr_slot_data)
1702 		return -ENOMEM;
1703 
1704 	for(i = 0; i < reg->hr_num_pages; i++) {
1705 		page = alloc_page(GFP_KERNEL);
1706 		if (!page)
1707 			return -ENOMEM;
1708 
1709 		reg->hr_slot_data[i] = page;
1710 
1711 		last_slot = i * spp;
1712 		raw = page_address(page);
1713 		for (j = 0;
1714 		     (j < spp) && ((j + last_slot) < reg->hr_blocks);
1715 		     j++) {
1716 			BUG_ON((j + last_slot) >= reg->hr_blocks);
1717 
1718 			slot = &reg->hr_slots[j + last_slot];
1719 			slot->ds_raw_block =
1720 				(struct o2hb_disk_heartbeat_block *) raw;
1721 
1722 			raw += reg->hr_block_bytes;
1723 		}
1724 	}
1725 
1726 	return 0;
1727 }
1728 
1729 /* Read in all the slots available and populate the tracking
1730  * structures so that we can start with a baseline idea of what's
1731  * there. */
1732 static int o2hb_populate_slot_data(struct o2hb_region *reg)
1733 {
1734 	int ret, i;
1735 	struct o2hb_disk_slot *slot;
1736 	struct o2hb_disk_heartbeat_block *hb_block;
1737 
1738 	ret = o2hb_read_slots(reg, 0, reg->hr_blocks);
1739 	if (ret)
1740 		goto out;
1741 
1742 	/* We only want to get an idea of the values initially in each
1743 	 * slot, so we do no verification - o2hb_check_slot will
1744 	 * actually determine if each configured slot is valid and
1745 	 * whether any values have changed. */
1746 	for(i = 0; i < reg->hr_blocks; i++) {
1747 		slot = &reg->hr_slots[i];
1748 		hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
1749 
1750 		/* Only fill the values that o2hb_check_slot uses to
1751 		 * determine changing slots */
1752 		slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
1753 		slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
1754 	}
1755 
1756 out:
1757 	return ret;
1758 }
1759 
1760 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
1761 static ssize_t o2hb_region_dev_store(struct config_item *item,
1762 				     const char *page,
1763 				     size_t count)
1764 {
1765 	struct o2hb_region *reg = to_o2hb_region(item);
1766 	struct task_struct *hb_task;
1767 	long fd;
1768 	int sectsize;
1769 	char *p = (char *)page;
1770 	struct fd f;
1771 	struct inode *inode;
1772 	ssize_t ret = -EINVAL;
1773 	int live_threshold;
1774 
1775 	if (reg->hr_bdev)
1776 		goto out;
1777 
1778 	/* We can't heartbeat without having had our node number
1779 	 * configured yet. */
1780 	if (o2nm_this_node() == O2NM_MAX_NODES)
1781 		goto out;
1782 
1783 	fd = simple_strtol(p, &p, 0);
1784 	if (!p || (*p && (*p != '\n')))
1785 		goto out;
1786 
1787 	if (fd < 0 || fd >= INT_MAX)
1788 		goto out;
1789 
1790 	f = fdget(fd);
1791 	if (f.file == NULL)
1792 		goto out;
1793 
1794 	if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
1795 	    reg->hr_block_bytes == 0)
1796 		goto out2;
1797 
1798 	inode = igrab(f.file->f_mapping->host);
1799 	if (inode == NULL)
1800 		goto out2;
1801 
1802 	if (!S_ISBLK(inode->i_mode))
1803 		goto out3;
1804 
1805 	reg->hr_bdev = I_BDEV(f.file->f_mapping->host);
1806 	ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL);
1807 	if (ret) {
1808 		reg->hr_bdev = NULL;
1809 		goto out3;
1810 	}
1811 	inode = NULL;
1812 
1813 	bdevname(reg->hr_bdev, reg->hr_dev_name);
1814 
1815 	sectsize = bdev_logical_block_size(reg->hr_bdev);
1816 	if (sectsize != reg->hr_block_bytes) {
1817 		mlog(ML_ERROR,
1818 		     "blocksize %u incorrect for device, expected %d",
1819 		     reg->hr_block_bytes, sectsize);
1820 		ret = -EINVAL;
1821 		goto out3;
1822 	}
1823 
1824 	o2hb_init_region_params(reg);
1825 
1826 	/* Generation of zero is invalid */
1827 	do {
1828 		get_random_bytes(&reg->hr_generation,
1829 				 sizeof(reg->hr_generation));
1830 	} while (reg->hr_generation == 0);
1831 
1832 	ret = o2hb_map_slot_data(reg);
1833 	if (ret) {
1834 		mlog_errno(ret);
1835 		goto out3;
1836 	}
1837 
1838 	ret = o2hb_populate_slot_data(reg);
1839 	if (ret) {
1840 		mlog_errno(ret);
1841 		goto out3;
1842 	}
1843 
1844 	INIT_DELAYED_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout);
1845 	INIT_DELAYED_WORK(&reg->hr_nego_timeout_work, o2hb_nego_timeout);
1846 
1847 	/*
1848 	 * A node is considered live after it has beat LIVE_THRESHOLD
1849 	 * times.  We're not steady until we've given them a chance
1850 	 * _after_ our first read.
1851 	 * The default threshold is bare minimum so as to limit the delay
1852 	 * during mounts. For global heartbeat, the threshold doubled for the
1853 	 * first region.
1854 	 */
1855 	live_threshold = O2HB_LIVE_THRESHOLD;
1856 	if (o2hb_global_heartbeat_active()) {
1857 		spin_lock(&o2hb_live_lock);
1858 		if (bitmap_weight(o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1)
1859 			live_threshold <<= 1;
1860 		spin_unlock(&o2hb_live_lock);
1861 	}
1862 	++live_threshold;
1863 	atomic_set(&reg->hr_steady_iterations, live_threshold);
1864 	/* unsteady_iterations is triple the steady_iterations */
1865 	atomic_set(&reg->hr_unsteady_iterations, (live_threshold * 3));
1866 
1867 	hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
1868 			      reg->hr_item.ci_name);
1869 	if (IS_ERR(hb_task)) {
1870 		ret = PTR_ERR(hb_task);
1871 		mlog_errno(ret);
1872 		goto out3;
1873 	}
1874 
1875 	spin_lock(&o2hb_live_lock);
1876 	reg->hr_task = hb_task;
1877 	spin_unlock(&o2hb_live_lock);
1878 
1879 	ret = wait_event_interruptible(o2hb_steady_queue,
1880 				atomic_read(&reg->hr_steady_iterations) == 0 ||
1881 				reg->hr_node_deleted);
1882 	if (ret) {
1883 		atomic_set(&reg->hr_steady_iterations, 0);
1884 		reg->hr_aborted_start = 1;
1885 	}
1886 
1887 	if (reg->hr_aborted_start) {
1888 		ret = -EIO;
1889 		goto out3;
1890 	}
1891 
1892 	if (reg->hr_node_deleted) {
1893 		ret = -EINVAL;
1894 		goto out3;
1895 	}
1896 
1897 	/* Ok, we were woken.  Make sure it wasn't by drop_item() */
1898 	spin_lock(&o2hb_live_lock);
1899 	hb_task = reg->hr_task;
1900 	if (o2hb_global_heartbeat_active())
1901 		set_bit(reg->hr_region_num, o2hb_live_region_bitmap);
1902 	spin_unlock(&o2hb_live_lock);
1903 
1904 	if (hb_task)
1905 		ret = count;
1906 	else
1907 		ret = -EIO;
1908 
1909 	if (hb_task && o2hb_global_heartbeat_active())
1910 		printk(KERN_NOTICE "o2hb: Heartbeat started on region %s (%s)\n",
1911 		       config_item_name(&reg->hr_item), reg->hr_dev_name);
1912 
1913 out3:
1914 	iput(inode);
1915 out2:
1916 	fdput(f);
1917 out:
1918 	if (ret < 0) {
1919 		if (reg->hr_bdev) {
1920 			blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1921 			reg->hr_bdev = NULL;
1922 		}
1923 	}
1924 	return ret;
1925 }
1926 
1927 static ssize_t o2hb_region_pid_show(struct config_item *item, char *page)
1928 {
1929 	struct o2hb_region *reg = to_o2hb_region(item);
1930 	pid_t pid = 0;
1931 
1932 	spin_lock(&o2hb_live_lock);
1933 	if (reg->hr_task)
1934 		pid = task_pid_nr(reg->hr_task);
1935 	spin_unlock(&o2hb_live_lock);
1936 
1937 	if (!pid)
1938 		return 0;
1939 
1940 	return sprintf(page, "%u\n", pid);
1941 }
1942 
1943 CONFIGFS_ATTR(o2hb_region_, block_bytes);
1944 CONFIGFS_ATTR(o2hb_region_, start_block);
1945 CONFIGFS_ATTR(o2hb_region_, blocks);
1946 CONFIGFS_ATTR(o2hb_region_, dev);
1947 CONFIGFS_ATTR_RO(o2hb_region_, pid);
1948 
1949 static struct configfs_attribute *o2hb_region_attrs[] = {
1950 	&o2hb_region_attr_block_bytes,
1951 	&o2hb_region_attr_start_block,
1952 	&o2hb_region_attr_blocks,
1953 	&o2hb_region_attr_dev,
1954 	&o2hb_region_attr_pid,
1955 	NULL,
1956 };
1957 
1958 static struct configfs_item_operations o2hb_region_item_ops = {
1959 	.release		= o2hb_region_release,
1960 };
1961 
1962 static const struct config_item_type o2hb_region_type = {
1963 	.ct_item_ops	= &o2hb_region_item_ops,
1964 	.ct_attrs	= o2hb_region_attrs,
1965 	.ct_owner	= THIS_MODULE,
1966 };
1967 
1968 /* heartbeat set */
1969 
1970 struct o2hb_heartbeat_group {
1971 	struct config_group hs_group;
1972 	/* some stuff? */
1973 };
1974 
1975 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
1976 {
1977 	return group ?
1978 		container_of(group, struct o2hb_heartbeat_group, hs_group)
1979 		: NULL;
1980 }
1981 
1982 static void o2hb_debug_region_init(struct o2hb_region *reg,
1983 				   struct dentry *parent)
1984 {
1985 	struct dentry *dir;
1986 
1987 	dir = debugfs_create_dir(config_item_name(&reg->hr_item), parent);
1988 	reg->hr_debug_dir = dir;
1989 
1990 	o2hb_debug_create(O2HB_DEBUG_LIVENODES, dir, &(reg->hr_db_livenodes),
1991 			  sizeof(*(reg->hr_db_livenodes)),
1992 			  O2HB_DB_TYPE_REGION_LIVENODES,
1993 			  sizeof(reg->hr_live_node_bitmap), O2NM_MAX_NODES,
1994 			  reg);
1995 
1996 	o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER, dir, &(reg->hr_db_regnum),
1997 			  sizeof(*(reg->hr_db_regnum)),
1998 			  O2HB_DB_TYPE_REGION_NUMBER, 0, O2NM_MAX_NODES, reg);
1999 
2000 	o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME, dir,
2001 			  &(reg->hr_db_elapsed_time),
2002 			  sizeof(*(reg->hr_db_elapsed_time)),
2003 			  O2HB_DB_TYPE_REGION_ELAPSED_TIME, 0, 0, reg);
2004 
2005 	o2hb_debug_create(O2HB_DEBUG_REGION_PINNED, dir, &(reg->hr_db_pinned),
2006 			  sizeof(*(reg->hr_db_pinned)),
2007 			  O2HB_DB_TYPE_REGION_PINNED, 0, 0, reg);
2008 
2009 }
2010 
2011 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
2012 							  const char *name)
2013 {
2014 	struct o2hb_region *reg = NULL;
2015 	int ret;
2016 
2017 	reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
2018 	if (reg == NULL)
2019 		return ERR_PTR(-ENOMEM);
2020 
2021 	if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) {
2022 		ret = -ENAMETOOLONG;
2023 		goto free;
2024 	}
2025 
2026 	spin_lock(&o2hb_live_lock);
2027 	reg->hr_region_num = 0;
2028 	if (o2hb_global_heartbeat_active()) {
2029 		reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap,
2030 							 O2NM_MAX_REGIONS);
2031 		if (reg->hr_region_num >= O2NM_MAX_REGIONS) {
2032 			spin_unlock(&o2hb_live_lock);
2033 			ret = -EFBIG;
2034 			goto free;
2035 		}
2036 		set_bit(reg->hr_region_num, o2hb_region_bitmap);
2037 	}
2038 	list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
2039 	spin_unlock(&o2hb_live_lock);
2040 
2041 	config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
2042 
2043 	/* this is the same way to generate msg key as dlm, for local heartbeat,
2044 	 * name is also the same, so make initial crc value different to avoid
2045 	 * message key conflict.
2046 	 */
2047 	reg->hr_key = crc32_le(reg->hr_region_num + O2NM_MAX_REGIONS,
2048 		name, strlen(name));
2049 	INIT_LIST_HEAD(&reg->hr_handler_list);
2050 	ret = o2net_register_handler(O2HB_NEGO_TIMEOUT_MSG, reg->hr_key,
2051 			sizeof(struct o2hb_nego_msg),
2052 			o2hb_nego_timeout_handler,
2053 			reg, NULL, &reg->hr_handler_list);
2054 	if (ret)
2055 		goto free;
2056 
2057 	ret = o2net_register_handler(O2HB_NEGO_APPROVE_MSG, reg->hr_key,
2058 			sizeof(struct o2hb_nego_msg),
2059 			o2hb_nego_approve_handler,
2060 			reg, NULL, &reg->hr_handler_list);
2061 	if (ret)
2062 		goto unregister_handler;
2063 
2064 	o2hb_debug_region_init(reg, o2hb_debug_dir);
2065 
2066 	return &reg->hr_item;
2067 
2068 unregister_handler:
2069 	o2net_unregister_handler_list(&reg->hr_handler_list);
2070 free:
2071 	kfree(reg);
2072 	return ERR_PTR(ret);
2073 }
2074 
2075 static void o2hb_heartbeat_group_drop_item(struct config_group *group,
2076 					   struct config_item *item)
2077 {
2078 	struct task_struct *hb_task;
2079 	struct o2hb_region *reg = to_o2hb_region(item);
2080 	int quorum_region = 0;
2081 
2082 	/* stop the thread when the user removes the region dir */
2083 	spin_lock(&o2hb_live_lock);
2084 	hb_task = reg->hr_task;
2085 	reg->hr_task = NULL;
2086 	reg->hr_item_dropped = 1;
2087 	spin_unlock(&o2hb_live_lock);
2088 
2089 	if (hb_task)
2090 		kthread_stop(hb_task);
2091 
2092 	if (o2hb_global_heartbeat_active()) {
2093 		spin_lock(&o2hb_live_lock);
2094 		clear_bit(reg->hr_region_num, o2hb_region_bitmap);
2095 		clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
2096 		if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
2097 			quorum_region = 1;
2098 		clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
2099 		spin_unlock(&o2hb_live_lock);
2100 		printk(KERN_NOTICE "o2hb: Heartbeat %s on region %s (%s)\n",
2101 		       ((atomic_read(&reg->hr_steady_iterations) == 0) ?
2102 			"stopped" : "start aborted"), config_item_name(item),
2103 		       reg->hr_dev_name);
2104 	}
2105 
2106 	/*
2107 	 * If we're racing a dev_write(), we need to wake them.  They will
2108 	 * check reg->hr_task
2109 	 */
2110 	if (atomic_read(&reg->hr_steady_iterations) != 0) {
2111 		reg->hr_aborted_start = 1;
2112 		atomic_set(&reg->hr_steady_iterations, 0);
2113 		wake_up(&o2hb_steady_queue);
2114 	}
2115 
2116 	config_item_put(item);
2117 
2118 	if (!o2hb_global_heartbeat_active() || !quorum_region)
2119 		return;
2120 
2121 	/*
2122 	 * If global heartbeat active and there are dependent users,
2123 	 * pin all regions if quorum region count <= CUT_OFF
2124 	 */
2125 	spin_lock(&o2hb_live_lock);
2126 
2127 	if (!o2hb_dependent_users)
2128 		goto unlock;
2129 
2130 	if (bitmap_weight(o2hb_quorum_region_bitmap,
2131 			   O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2132 		o2hb_region_pin(NULL);
2133 
2134 unlock:
2135 	spin_unlock(&o2hb_live_lock);
2136 }
2137 
2138 static ssize_t o2hb_heartbeat_group_dead_threshold_show(struct config_item *item,
2139 		char *page)
2140 {
2141 	return sprintf(page, "%u\n", o2hb_dead_threshold);
2142 }
2143 
2144 static ssize_t o2hb_heartbeat_group_dead_threshold_store(struct config_item *item,
2145 		const char *page, size_t count)
2146 {
2147 	unsigned long tmp;
2148 	char *p = (char *)page;
2149 
2150 	tmp = simple_strtoul(p, &p, 10);
2151 	if (!p || (*p && (*p != '\n')))
2152                 return -EINVAL;
2153 
2154 	/* this will validate ranges for us. */
2155 	o2hb_dead_threshold_set((unsigned int) tmp);
2156 
2157 	return count;
2158 }
2159 
2160 static ssize_t o2hb_heartbeat_group_mode_show(struct config_item *item,
2161 		char *page)
2162 {
2163 	return sprintf(page, "%s\n",
2164 		       o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]);
2165 }
2166 
2167 static ssize_t o2hb_heartbeat_group_mode_store(struct config_item *item,
2168 		const char *page, size_t count)
2169 {
2170 	unsigned int i;
2171 	int ret;
2172 	size_t len;
2173 
2174 	len = (page[count - 1] == '\n') ? count - 1 : count;
2175 	if (!len)
2176 		return -EINVAL;
2177 
2178 	for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) {
2179 		if (strncasecmp(page, o2hb_heartbeat_mode_desc[i], len))
2180 			continue;
2181 
2182 		ret = o2hb_global_heartbeat_mode_set(i);
2183 		if (!ret)
2184 			printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n",
2185 			       o2hb_heartbeat_mode_desc[i]);
2186 		return count;
2187 	}
2188 
2189 	return -EINVAL;
2190 
2191 }
2192 
2193 CONFIGFS_ATTR(o2hb_heartbeat_group_, dead_threshold);
2194 CONFIGFS_ATTR(o2hb_heartbeat_group_, mode);
2195 
2196 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
2197 	&o2hb_heartbeat_group_attr_dead_threshold,
2198 	&o2hb_heartbeat_group_attr_mode,
2199 	NULL,
2200 };
2201 
2202 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
2203 	.make_item	= o2hb_heartbeat_group_make_item,
2204 	.drop_item	= o2hb_heartbeat_group_drop_item,
2205 };
2206 
2207 static const struct config_item_type o2hb_heartbeat_group_type = {
2208 	.ct_group_ops	= &o2hb_heartbeat_group_group_ops,
2209 	.ct_attrs	= o2hb_heartbeat_group_attrs,
2210 	.ct_owner	= THIS_MODULE,
2211 };
2212 
2213 /* this is just here to avoid touching group in heartbeat.h which the
2214  * entire damn world #includes */
2215 struct config_group *o2hb_alloc_hb_set(void)
2216 {
2217 	struct o2hb_heartbeat_group *hs = NULL;
2218 	struct config_group *ret = NULL;
2219 
2220 	hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
2221 	if (hs == NULL)
2222 		goto out;
2223 
2224 	config_group_init_type_name(&hs->hs_group, "heartbeat",
2225 				    &o2hb_heartbeat_group_type);
2226 
2227 	ret = &hs->hs_group;
2228 out:
2229 	if (ret == NULL)
2230 		kfree(hs);
2231 	return ret;
2232 }
2233 
2234 void o2hb_free_hb_set(struct config_group *group)
2235 {
2236 	struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
2237 	kfree(hs);
2238 }
2239 
2240 /* hb callback registration and issuing */
2241 
2242 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
2243 {
2244 	if (type == O2HB_NUM_CB)
2245 		return ERR_PTR(-EINVAL);
2246 
2247 	return &o2hb_callbacks[type];
2248 }
2249 
2250 void o2hb_setup_callback(struct o2hb_callback_func *hc,
2251 			 enum o2hb_callback_type type,
2252 			 o2hb_cb_func *func,
2253 			 void *data,
2254 			 int priority)
2255 {
2256 	INIT_LIST_HEAD(&hc->hc_item);
2257 	hc->hc_func = func;
2258 	hc->hc_data = data;
2259 	hc->hc_priority = priority;
2260 	hc->hc_type = type;
2261 	hc->hc_magic = O2HB_CB_MAGIC;
2262 }
2263 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
2264 
2265 /*
2266  * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2267  * In global heartbeat mode, region_uuid passed is NULL.
2268  *
2269  * In local, we only pin the matching region. In global we pin all the active
2270  * regions.
2271  */
2272 static int o2hb_region_pin(const char *region_uuid)
2273 {
2274 	int ret = 0, found = 0;
2275 	struct o2hb_region *reg;
2276 	char *uuid;
2277 
2278 	assert_spin_locked(&o2hb_live_lock);
2279 
2280 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2281 		if (reg->hr_item_dropped)
2282 			continue;
2283 
2284 		uuid = config_item_name(&reg->hr_item);
2285 
2286 		/* local heartbeat */
2287 		if (region_uuid) {
2288 			if (strcmp(region_uuid, uuid))
2289 				continue;
2290 			found = 1;
2291 		}
2292 
2293 		if (reg->hr_item_pinned || reg->hr_item_dropped)
2294 			goto skip_pin;
2295 
2296 		/* Ignore ENOENT only for local hb (userdlm domain) */
2297 		ret = o2nm_depend_item(&reg->hr_item);
2298 		if (!ret) {
2299 			mlog(ML_CLUSTER, "Pin region %s\n", uuid);
2300 			reg->hr_item_pinned = 1;
2301 		} else {
2302 			if (ret == -ENOENT && found)
2303 				ret = 0;
2304 			else {
2305 				mlog(ML_ERROR, "Pin region %s fails with %d\n",
2306 				     uuid, ret);
2307 				break;
2308 			}
2309 		}
2310 skip_pin:
2311 		if (found)
2312 			break;
2313 	}
2314 
2315 	return ret;
2316 }
2317 
2318 /*
2319  * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2320  * In global heartbeat mode, region_uuid passed is NULL.
2321  *
2322  * In local, we only unpin the matching region. In global we unpin all the
2323  * active regions.
2324  */
2325 static void o2hb_region_unpin(const char *region_uuid)
2326 {
2327 	struct o2hb_region *reg;
2328 	char *uuid;
2329 	int found = 0;
2330 
2331 	assert_spin_locked(&o2hb_live_lock);
2332 
2333 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2334 		if (reg->hr_item_dropped)
2335 			continue;
2336 
2337 		uuid = config_item_name(&reg->hr_item);
2338 		if (region_uuid) {
2339 			if (strcmp(region_uuid, uuid))
2340 				continue;
2341 			found = 1;
2342 		}
2343 
2344 		if (reg->hr_item_pinned) {
2345 			mlog(ML_CLUSTER, "Unpin region %s\n", uuid);
2346 			o2nm_undepend_item(&reg->hr_item);
2347 			reg->hr_item_pinned = 0;
2348 		}
2349 		if (found)
2350 			break;
2351 	}
2352 }
2353 
2354 static int o2hb_region_inc_user(const char *region_uuid)
2355 {
2356 	int ret = 0;
2357 
2358 	spin_lock(&o2hb_live_lock);
2359 
2360 	/* local heartbeat */
2361 	if (!o2hb_global_heartbeat_active()) {
2362 	    ret = o2hb_region_pin(region_uuid);
2363 	    goto unlock;
2364 	}
2365 
2366 	/*
2367 	 * if global heartbeat active and this is the first dependent user,
2368 	 * pin all regions if quorum region count <= CUT_OFF
2369 	 */
2370 	o2hb_dependent_users++;
2371 	if (o2hb_dependent_users > 1)
2372 		goto unlock;
2373 
2374 	if (bitmap_weight(o2hb_quorum_region_bitmap,
2375 			   O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2376 		ret = o2hb_region_pin(NULL);
2377 
2378 unlock:
2379 	spin_unlock(&o2hb_live_lock);
2380 	return ret;
2381 }
2382 
2383 static void o2hb_region_dec_user(const char *region_uuid)
2384 {
2385 	spin_lock(&o2hb_live_lock);
2386 
2387 	/* local heartbeat */
2388 	if (!o2hb_global_heartbeat_active()) {
2389 	    o2hb_region_unpin(region_uuid);
2390 	    goto unlock;
2391 	}
2392 
2393 	/*
2394 	 * if global heartbeat active and there are no dependent users,
2395 	 * unpin all quorum regions
2396 	 */
2397 	o2hb_dependent_users--;
2398 	if (!o2hb_dependent_users)
2399 		o2hb_region_unpin(NULL);
2400 
2401 unlock:
2402 	spin_unlock(&o2hb_live_lock);
2403 }
2404 
2405 int o2hb_register_callback(const char *region_uuid,
2406 			   struct o2hb_callback_func *hc)
2407 {
2408 	struct o2hb_callback_func *f;
2409 	struct o2hb_callback *hbcall;
2410 	int ret;
2411 
2412 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2413 	BUG_ON(!list_empty(&hc->hc_item));
2414 
2415 	hbcall = hbcall_from_type(hc->hc_type);
2416 	if (IS_ERR(hbcall)) {
2417 		ret = PTR_ERR(hbcall);
2418 		goto out;
2419 	}
2420 
2421 	if (region_uuid) {
2422 		ret = o2hb_region_inc_user(region_uuid);
2423 		if (ret) {
2424 			mlog_errno(ret);
2425 			goto out;
2426 		}
2427 	}
2428 
2429 	down_write(&o2hb_callback_sem);
2430 
2431 	list_for_each_entry(f, &hbcall->list, hc_item) {
2432 		if (hc->hc_priority < f->hc_priority) {
2433 			list_add_tail(&hc->hc_item, &f->hc_item);
2434 			break;
2435 		}
2436 	}
2437 	if (list_empty(&hc->hc_item))
2438 		list_add_tail(&hc->hc_item, &hbcall->list);
2439 
2440 	up_write(&o2hb_callback_sem);
2441 	ret = 0;
2442 out:
2443 	mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n",
2444 	     ret, __builtin_return_address(0), hc);
2445 	return ret;
2446 }
2447 EXPORT_SYMBOL_GPL(o2hb_register_callback);
2448 
2449 void o2hb_unregister_callback(const char *region_uuid,
2450 			      struct o2hb_callback_func *hc)
2451 {
2452 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2453 
2454 	mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n",
2455 	     __builtin_return_address(0), hc);
2456 
2457 	/* XXX Can this happen _with_ a region reference? */
2458 	if (list_empty(&hc->hc_item))
2459 		return;
2460 
2461 	if (region_uuid)
2462 		o2hb_region_dec_user(region_uuid);
2463 
2464 	down_write(&o2hb_callback_sem);
2465 
2466 	list_del_init(&hc->hc_item);
2467 
2468 	up_write(&o2hb_callback_sem);
2469 }
2470 EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
2471 
2472 int o2hb_check_node_heartbeating_no_sem(u8 node_num)
2473 {
2474 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2475 
2476 	spin_lock(&o2hb_live_lock);
2477 	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2478 	spin_unlock(&o2hb_live_lock);
2479 	if (!test_bit(node_num, testing_map)) {
2480 		mlog(ML_HEARTBEAT,
2481 		     "node (%u) does not have heartbeating enabled.\n",
2482 		     node_num);
2483 		return 0;
2484 	}
2485 
2486 	return 1;
2487 }
2488 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_no_sem);
2489 
2490 int o2hb_check_node_heartbeating_from_callback(u8 node_num)
2491 {
2492 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2493 
2494 	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2495 	if (!test_bit(node_num, testing_map)) {
2496 		mlog(ML_HEARTBEAT,
2497 		     "node (%u) does not have heartbeating enabled.\n",
2498 		     node_num);
2499 		return 0;
2500 	}
2501 
2502 	return 1;
2503 }
2504 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
2505 
2506 /*
2507  * this is just a hack until we get the plumbing which flips file systems
2508  * read only and drops the hb ref instead of killing the node dead.
2509  */
2510 void o2hb_stop_all_regions(void)
2511 {
2512 	struct o2hb_region *reg;
2513 
2514 	mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
2515 
2516 	spin_lock(&o2hb_live_lock);
2517 
2518 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
2519 		reg->hr_unclean_stop = 1;
2520 
2521 	spin_unlock(&o2hb_live_lock);
2522 }
2523 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
2524 
2525 int o2hb_get_all_regions(char *region_uuids, u8 max_regions)
2526 {
2527 	struct o2hb_region *reg;
2528 	int numregs = 0;
2529 	char *p;
2530 
2531 	spin_lock(&o2hb_live_lock);
2532 
2533 	p = region_uuids;
2534 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2535 		if (reg->hr_item_dropped)
2536 			continue;
2537 
2538 		mlog(0, "Region: %s\n", config_item_name(&reg->hr_item));
2539 		if (numregs < max_regions) {
2540 			memcpy(p, config_item_name(&reg->hr_item),
2541 			       O2HB_MAX_REGION_NAME_LEN);
2542 			p += O2HB_MAX_REGION_NAME_LEN;
2543 		}
2544 		numregs++;
2545 	}
2546 
2547 	spin_unlock(&o2hb_live_lock);
2548 
2549 	return numregs;
2550 }
2551 EXPORT_SYMBOL_GPL(o2hb_get_all_regions);
2552 
2553 int o2hb_global_heartbeat_active(void)
2554 {
2555 	return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL);
2556 }
2557 EXPORT_SYMBOL(o2hb_global_heartbeat_active);
2558