xref: /openbmc/linux/fs/ocfs2/cluster/heartbeat.c (revision 5d0e4d78)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * Copyright (C) 2004, 2005 Oracle.  All rights reserved.
5  *
6  * This program is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU General Public
8  * License as published by the Free Software Foundation; either
9  * version 2 of the License, or (at your option) any later version.
10  *
11  * This program is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public
17  * License along with this program; if not, write to the
18  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19  * Boston, MA 021110-1307, USA.
20  */
21 
22 #include <linux/kernel.h>
23 #include <linux/sched.h>
24 #include <linux/jiffies.h>
25 #include <linux/module.h>
26 #include <linux/fs.h>
27 #include <linux/bio.h>
28 #include <linux/blkdev.h>
29 #include <linux/delay.h>
30 #include <linux/file.h>
31 #include <linux/kthread.h>
32 #include <linux/configfs.h>
33 #include <linux/random.h>
34 #include <linux/crc32.h>
35 #include <linux/time.h>
36 #include <linux/debugfs.h>
37 #include <linux/slab.h>
38 #include <linux/bitmap.h>
39 #include <linux/ktime.h>
40 #include "heartbeat.h"
41 #include "tcp.h"
42 #include "nodemanager.h"
43 #include "quorum.h"
44 
45 #include "masklog.h"
46 
47 
48 /*
49  * The first heartbeat pass had one global thread that would serialize all hb
50  * callback calls.  This global serializing sem should only be removed once
51  * we've made sure that all callees can deal with being called concurrently
52  * from multiple hb region threads.
53  */
54 static DECLARE_RWSEM(o2hb_callback_sem);
55 
56 /*
57  * multiple hb threads are watching multiple regions.  A node is live
58  * whenever any of the threads sees activity from the node in its region.
59  */
60 static DEFINE_SPINLOCK(o2hb_live_lock);
61 static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
62 static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
63 static LIST_HEAD(o2hb_node_events);
64 static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
65 
66 /*
67  * In global heartbeat, we maintain a series of region bitmaps.
68  * 	- o2hb_region_bitmap allows us to limit the region number to max region.
69  * 	- o2hb_live_region_bitmap tracks live regions (seen steady iterations).
70  * 	- o2hb_quorum_region_bitmap tracks live regions that have seen all nodes
71  * 		heartbeat on it.
72  * 	- o2hb_failed_region_bitmap tracks the regions that have seen io timeouts.
73  */
74 static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
75 static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
76 static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
77 static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
78 
79 #define O2HB_DB_TYPE_LIVENODES		0
80 #define O2HB_DB_TYPE_LIVEREGIONS	1
81 #define O2HB_DB_TYPE_QUORUMREGIONS	2
82 #define O2HB_DB_TYPE_FAILEDREGIONS	3
83 #define O2HB_DB_TYPE_REGION_LIVENODES	4
84 #define O2HB_DB_TYPE_REGION_NUMBER	5
85 #define O2HB_DB_TYPE_REGION_ELAPSED_TIME	6
86 #define O2HB_DB_TYPE_REGION_PINNED	7
87 struct o2hb_debug_buf {
88 	int db_type;
89 	int db_size;
90 	int db_len;
91 	void *db_data;
92 };
93 
94 static struct o2hb_debug_buf *o2hb_db_livenodes;
95 static struct o2hb_debug_buf *o2hb_db_liveregions;
96 static struct o2hb_debug_buf *o2hb_db_quorumregions;
97 static struct o2hb_debug_buf *o2hb_db_failedregions;
98 
99 #define O2HB_DEBUG_DIR			"o2hb"
100 #define O2HB_DEBUG_LIVENODES		"livenodes"
101 #define O2HB_DEBUG_LIVEREGIONS		"live_regions"
102 #define O2HB_DEBUG_QUORUMREGIONS	"quorum_regions"
103 #define O2HB_DEBUG_FAILEDREGIONS	"failed_regions"
104 #define O2HB_DEBUG_REGION_NUMBER	"num"
105 #define O2HB_DEBUG_REGION_ELAPSED_TIME	"elapsed_time_in_ms"
106 #define O2HB_DEBUG_REGION_PINNED	"pinned"
107 
108 static struct dentry *o2hb_debug_dir;
109 static struct dentry *o2hb_debug_livenodes;
110 static struct dentry *o2hb_debug_liveregions;
111 static struct dentry *o2hb_debug_quorumregions;
112 static struct dentry *o2hb_debug_failedregions;
113 
114 static LIST_HEAD(o2hb_all_regions);
115 
116 static struct o2hb_callback {
117 	struct list_head list;
118 } o2hb_callbacks[O2HB_NUM_CB];
119 
120 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
121 
122 #define O2HB_DEFAULT_BLOCK_BITS       9
123 
124 enum o2hb_heartbeat_modes {
125 	O2HB_HEARTBEAT_LOCAL		= 0,
126 	O2HB_HEARTBEAT_GLOBAL,
127 	O2HB_HEARTBEAT_NUM_MODES,
128 };
129 
130 char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = {
131 		"local",	/* O2HB_HEARTBEAT_LOCAL */
132 		"global",	/* O2HB_HEARTBEAT_GLOBAL */
133 };
134 
135 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
136 unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL;
137 
138 /*
139  * o2hb_dependent_users tracks the number of registered callbacks that depend
140  * on heartbeat. o2net and o2dlm are two entities that register this callback.
141  * However only o2dlm depends on the heartbeat. It does not want the heartbeat
142  * to stop while a dlm domain is still active.
143  */
144 unsigned int o2hb_dependent_users;
145 
146 /*
147  * In global heartbeat mode, all regions are pinned if there are one or more
148  * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All
149  * regions are unpinned if the region count exceeds the cut off or the number
150  * of dependent users falls to zero.
151  */
152 #define O2HB_PIN_CUT_OFF		3
153 
154 /*
155  * In local heartbeat mode, we assume the dlm domain name to be the same as
156  * region uuid. This is true for domains created for the file system but not
157  * necessarily true for userdlm domains. This is a known limitation.
158  *
159  * In global heartbeat mode, we pin/unpin all o2hb regions. This solution
160  * works for both file system and userdlm domains.
161  */
162 static int o2hb_region_pin(const char *region_uuid);
163 static void o2hb_region_unpin(const char *region_uuid);
164 
165 /* Only sets a new threshold if there are no active regions.
166  *
167  * No locking or otherwise interesting code is required for reading
168  * o2hb_dead_threshold as it can't change once regions are active and
169  * it's not interesting to anyone until then anyway. */
170 static void o2hb_dead_threshold_set(unsigned int threshold)
171 {
172 	if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
173 		spin_lock(&o2hb_live_lock);
174 		if (list_empty(&o2hb_all_regions))
175 			o2hb_dead_threshold = threshold;
176 		spin_unlock(&o2hb_live_lock);
177 	}
178 }
179 
180 static int o2hb_global_heartbeat_mode_set(unsigned int hb_mode)
181 {
182 	int ret = -1;
183 
184 	if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) {
185 		spin_lock(&o2hb_live_lock);
186 		if (list_empty(&o2hb_all_regions)) {
187 			o2hb_heartbeat_mode = hb_mode;
188 			ret = 0;
189 		}
190 		spin_unlock(&o2hb_live_lock);
191 	}
192 
193 	return ret;
194 }
195 
196 struct o2hb_node_event {
197 	struct list_head        hn_item;
198 	enum o2hb_callback_type hn_event_type;
199 	struct o2nm_node        *hn_node;
200 	int                     hn_node_num;
201 };
202 
203 struct o2hb_disk_slot {
204 	struct o2hb_disk_heartbeat_block *ds_raw_block;
205 	u8			ds_node_num;
206 	u64			ds_last_time;
207 	u64			ds_last_generation;
208 	u16			ds_equal_samples;
209 	u16			ds_changed_samples;
210 	struct list_head	ds_live_item;
211 };
212 
213 /* each thread owns a region.. when we're asked to tear down the region
214  * we ask the thread to stop, who cleans up the region */
215 struct o2hb_region {
216 	struct config_item	hr_item;
217 
218 	struct list_head	hr_all_item;
219 	unsigned		hr_unclean_stop:1,
220 				hr_aborted_start:1,
221 				hr_item_pinned:1,
222 				hr_item_dropped:1,
223 				hr_node_deleted:1;
224 
225 	/* protected by the hr_callback_sem */
226 	struct task_struct 	*hr_task;
227 
228 	unsigned int		hr_blocks;
229 	unsigned long long	hr_start_block;
230 
231 	unsigned int		hr_block_bits;
232 	unsigned int		hr_block_bytes;
233 
234 	unsigned int		hr_slots_per_page;
235 	unsigned int		hr_num_pages;
236 
237 	struct page             **hr_slot_data;
238 	struct block_device	*hr_bdev;
239 	struct o2hb_disk_slot	*hr_slots;
240 
241 	/* live node map of this region */
242 	unsigned long		hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
243 	unsigned int		hr_region_num;
244 
245 	struct dentry		*hr_debug_dir;
246 	struct dentry		*hr_debug_livenodes;
247 	struct dentry		*hr_debug_regnum;
248 	struct dentry		*hr_debug_elapsed_time;
249 	struct dentry		*hr_debug_pinned;
250 	struct o2hb_debug_buf	*hr_db_livenodes;
251 	struct o2hb_debug_buf	*hr_db_regnum;
252 	struct o2hb_debug_buf	*hr_db_elapsed_time;
253 	struct o2hb_debug_buf	*hr_db_pinned;
254 
255 	/* let the person setting up hb wait for it to return until it
256 	 * has reached a 'steady' state.  This will be fixed when we have
257 	 * a more complete api that doesn't lead to this sort of fragility. */
258 	atomic_t		hr_steady_iterations;
259 
260 	/* terminate o2hb thread if it does not reach steady state
261 	 * (hr_steady_iterations == 0) within hr_unsteady_iterations */
262 	atomic_t		hr_unsteady_iterations;
263 
264 	char			hr_dev_name[BDEVNAME_SIZE];
265 
266 	unsigned int		hr_timeout_ms;
267 
268 	/* randomized as the region goes up and down so that a node
269 	 * recognizes a node going up and down in one iteration */
270 	u64			hr_generation;
271 
272 	struct delayed_work	hr_write_timeout_work;
273 	unsigned long		hr_last_timeout_start;
274 
275 	/* negotiate timer, used to negotiate extending hb timeout. */
276 	struct delayed_work	hr_nego_timeout_work;
277 	unsigned long		hr_nego_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
278 
279 	/* Used during o2hb_check_slot to hold a copy of the block
280 	 * being checked because we temporarily have to zero out the
281 	 * crc field. */
282 	struct o2hb_disk_heartbeat_block *hr_tmp_block;
283 
284 	/* Message key for negotiate timeout message. */
285 	unsigned int		hr_key;
286 	struct list_head	hr_handler_list;
287 
288 	/* last hb status, 0 for success, other value for error. */
289 	int			hr_last_hb_status;
290 };
291 
292 struct o2hb_bio_wait_ctxt {
293 	atomic_t          wc_num_reqs;
294 	struct completion wc_io_complete;
295 	int               wc_error;
296 };
297 
298 #define O2HB_NEGO_TIMEOUT_MS (O2HB_MAX_WRITE_TIMEOUT_MS/2)
299 
300 enum {
301 	O2HB_NEGO_TIMEOUT_MSG = 1,
302 	O2HB_NEGO_APPROVE_MSG = 2,
303 };
304 
305 struct o2hb_nego_msg {
306 	u8 node_num;
307 };
308 
309 static void o2hb_write_timeout(struct work_struct *work)
310 {
311 	int failed, quorum;
312 	struct o2hb_region *reg =
313 		container_of(work, struct o2hb_region,
314 			     hr_write_timeout_work.work);
315 
316 	mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
317 	     "milliseconds\n", reg->hr_dev_name,
318 	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
319 
320 	if (o2hb_global_heartbeat_active()) {
321 		spin_lock(&o2hb_live_lock);
322 		if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
323 			set_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
324 		failed = bitmap_weight(o2hb_failed_region_bitmap,
325 					O2NM_MAX_REGIONS);
326 		quorum = bitmap_weight(o2hb_quorum_region_bitmap,
327 					O2NM_MAX_REGIONS);
328 		spin_unlock(&o2hb_live_lock);
329 
330 		mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n",
331 		     quorum, failed);
332 
333 		/*
334 		 * Fence if the number of failed regions >= half the number
335 		 * of  quorum regions
336 		 */
337 		if ((failed << 1) < quorum)
338 			return;
339 	}
340 
341 	o2quo_disk_timeout();
342 }
343 
344 static void o2hb_arm_timeout(struct o2hb_region *reg)
345 {
346 	/* Arm writeout only after thread reaches steady state */
347 	if (atomic_read(&reg->hr_steady_iterations) != 0)
348 		return;
349 
350 	mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n",
351 	     O2HB_MAX_WRITE_TIMEOUT_MS);
352 
353 	if (o2hb_global_heartbeat_active()) {
354 		spin_lock(&o2hb_live_lock);
355 		clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
356 		spin_unlock(&o2hb_live_lock);
357 	}
358 	cancel_delayed_work(&reg->hr_write_timeout_work);
359 	schedule_delayed_work(&reg->hr_write_timeout_work,
360 			      msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
361 
362 	cancel_delayed_work(&reg->hr_nego_timeout_work);
363 	/* negotiate timeout must be less than write timeout. */
364 	schedule_delayed_work(&reg->hr_nego_timeout_work,
365 			      msecs_to_jiffies(O2HB_NEGO_TIMEOUT_MS));
366 	memset(reg->hr_nego_node_bitmap, 0, sizeof(reg->hr_nego_node_bitmap));
367 }
368 
369 static void o2hb_disarm_timeout(struct o2hb_region *reg)
370 {
371 	cancel_delayed_work_sync(&reg->hr_write_timeout_work);
372 	cancel_delayed_work_sync(&reg->hr_nego_timeout_work);
373 }
374 
375 static int o2hb_send_nego_msg(int key, int type, u8 target)
376 {
377 	struct o2hb_nego_msg msg;
378 	int status, ret;
379 
380 	msg.node_num = o2nm_this_node();
381 again:
382 	ret = o2net_send_message(type, key, &msg, sizeof(msg),
383 			target, &status);
384 
385 	if (ret == -EAGAIN || ret == -ENOMEM) {
386 		msleep(100);
387 		goto again;
388 	}
389 
390 	return ret;
391 }
392 
393 static void o2hb_nego_timeout(struct work_struct *work)
394 {
395 	unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
396 	int master_node, i, ret;
397 	struct o2hb_region *reg;
398 
399 	reg = container_of(work, struct o2hb_region, hr_nego_timeout_work.work);
400 	/* don't negotiate timeout if last hb failed since it is very
401 	 * possible io failed. Should let write timeout fence self.
402 	 */
403 	if (reg->hr_last_hb_status)
404 		return;
405 
406 	o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
407 	/* lowest node as master node to make negotiate decision. */
408 	master_node = find_next_bit(live_node_bitmap, O2NM_MAX_NODES, 0);
409 
410 	if (master_node == o2nm_this_node()) {
411 		if (!test_bit(master_node, reg->hr_nego_node_bitmap)) {
412 			printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s).\n",
413 				o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000,
414 				config_item_name(&reg->hr_item), reg->hr_dev_name);
415 			set_bit(master_node, reg->hr_nego_node_bitmap);
416 		}
417 		if (memcmp(reg->hr_nego_node_bitmap, live_node_bitmap,
418 				sizeof(reg->hr_nego_node_bitmap))) {
419 			/* check negotiate bitmap every second to do timeout
420 			 * approve decision.
421 			 */
422 			schedule_delayed_work(&reg->hr_nego_timeout_work,
423 				msecs_to_jiffies(1000));
424 
425 			return;
426 		}
427 
428 		printk(KERN_NOTICE "o2hb: all nodes hb write hung, maybe region %s (%s) is down.\n",
429 			config_item_name(&reg->hr_item), reg->hr_dev_name);
430 		/* approve negotiate timeout request. */
431 		o2hb_arm_timeout(reg);
432 
433 		i = -1;
434 		while ((i = find_next_bit(live_node_bitmap,
435 				O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
436 			if (i == master_node)
437 				continue;
438 
439 			mlog(ML_HEARTBEAT, "send NEGO_APPROVE msg to node %d\n", i);
440 			ret = o2hb_send_nego_msg(reg->hr_key,
441 					O2HB_NEGO_APPROVE_MSG, i);
442 			if (ret)
443 				mlog(ML_ERROR, "send NEGO_APPROVE msg to node %d fail %d\n",
444 					i, ret);
445 		}
446 	} else {
447 		/* negotiate timeout with master node. */
448 		printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s), negotiate timeout with node %d.\n",
449 			o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000, config_item_name(&reg->hr_item),
450 			reg->hr_dev_name, master_node);
451 		ret = o2hb_send_nego_msg(reg->hr_key, O2HB_NEGO_TIMEOUT_MSG,
452 				master_node);
453 		if (ret)
454 			mlog(ML_ERROR, "send NEGO_TIMEOUT msg to node %d fail %d\n",
455 				master_node, ret);
456 	}
457 }
458 
459 static int o2hb_nego_timeout_handler(struct o2net_msg *msg, u32 len, void *data,
460 				void **ret_data)
461 {
462 	struct o2hb_region *reg = data;
463 	struct o2hb_nego_msg *nego_msg;
464 
465 	nego_msg = (struct o2hb_nego_msg *)msg->buf;
466 	printk(KERN_NOTICE "o2hb: receive negotiate timeout message from node %d on region %s (%s).\n",
467 		nego_msg->node_num, config_item_name(&reg->hr_item), reg->hr_dev_name);
468 	if (nego_msg->node_num < O2NM_MAX_NODES)
469 		set_bit(nego_msg->node_num, reg->hr_nego_node_bitmap);
470 	else
471 		mlog(ML_ERROR, "got nego timeout message from bad node.\n");
472 
473 	return 0;
474 }
475 
476 static int o2hb_nego_approve_handler(struct o2net_msg *msg, u32 len, void *data,
477 				void **ret_data)
478 {
479 	struct o2hb_region *reg = data;
480 
481 	printk(KERN_NOTICE "o2hb: negotiate timeout approved by master node on region %s (%s).\n",
482 		config_item_name(&reg->hr_item), reg->hr_dev_name);
483 	o2hb_arm_timeout(reg);
484 	return 0;
485 }
486 
487 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc)
488 {
489 	atomic_set(&wc->wc_num_reqs, 1);
490 	init_completion(&wc->wc_io_complete);
491 	wc->wc_error = 0;
492 }
493 
494 /* Used in error paths too */
495 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
496 				     unsigned int num)
497 {
498 	/* sadly atomic_sub_and_test() isn't available on all platforms.  The
499 	 * good news is that the fast path only completes one at a time */
500 	while(num--) {
501 		if (atomic_dec_and_test(&wc->wc_num_reqs)) {
502 			BUG_ON(num > 0);
503 			complete(&wc->wc_io_complete);
504 		}
505 	}
506 }
507 
508 static void o2hb_wait_on_io(struct o2hb_region *reg,
509 			    struct o2hb_bio_wait_ctxt *wc)
510 {
511 	o2hb_bio_wait_dec(wc, 1);
512 	wait_for_completion(&wc->wc_io_complete);
513 }
514 
515 static void o2hb_bio_end_io(struct bio *bio)
516 {
517 	struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
518 
519 	if (bio->bi_status) {
520 		mlog(ML_ERROR, "IO Error %d\n", bio->bi_status);
521 		wc->wc_error = blk_status_to_errno(bio->bi_status);
522 	}
523 
524 	o2hb_bio_wait_dec(wc, 1);
525 	bio_put(bio);
526 }
527 
528 /* Setup a Bio to cover I/O against num_slots slots starting at
529  * start_slot. */
530 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
531 				      struct o2hb_bio_wait_ctxt *wc,
532 				      unsigned int *current_slot,
533 				      unsigned int max_slots, int op,
534 				      int op_flags)
535 {
536 	int len, current_page;
537 	unsigned int vec_len, vec_start;
538 	unsigned int bits = reg->hr_block_bits;
539 	unsigned int spp = reg->hr_slots_per_page;
540 	unsigned int cs = *current_slot;
541 	struct bio *bio;
542 	struct page *page;
543 
544 	/* Testing has shown this allocation to take long enough under
545 	 * GFP_KERNEL that the local node can get fenced. It would be
546 	 * nicest if we could pre-allocate these bios and avoid this
547 	 * all together. */
548 	bio = bio_alloc(GFP_ATOMIC, 16);
549 	if (!bio) {
550 		mlog(ML_ERROR, "Could not alloc slots BIO!\n");
551 		bio = ERR_PTR(-ENOMEM);
552 		goto bail;
553 	}
554 
555 	/* Must put everything in 512 byte sectors for the bio... */
556 	bio->bi_iter.bi_sector = (reg->hr_start_block + cs) << (bits - 9);
557 	bio->bi_bdev = reg->hr_bdev;
558 	bio->bi_private = wc;
559 	bio->bi_end_io = o2hb_bio_end_io;
560 	bio_set_op_attrs(bio, op, op_flags);
561 
562 	vec_start = (cs << bits) % PAGE_SIZE;
563 	while(cs < max_slots) {
564 		current_page = cs / spp;
565 		page = reg->hr_slot_data[current_page];
566 
567 		vec_len = min(PAGE_SIZE - vec_start,
568 			      (max_slots-cs) * (PAGE_SIZE/spp) );
569 
570 		mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
571 		     current_page, vec_len, vec_start);
572 
573 		len = bio_add_page(bio, page, vec_len, vec_start);
574 		if (len != vec_len) break;
575 
576 		cs += vec_len / (PAGE_SIZE/spp);
577 		vec_start = 0;
578 	}
579 
580 bail:
581 	*current_slot = cs;
582 	return bio;
583 }
584 
585 static int o2hb_read_slots(struct o2hb_region *reg,
586 			   unsigned int max_slots)
587 {
588 	unsigned int current_slot=0;
589 	int status;
590 	struct o2hb_bio_wait_ctxt wc;
591 	struct bio *bio;
592 
593 	o2hb_bio_wait_init(&wc);
594 
595 	while(current_slot < max_slots) {
596 		bio = o2hb_setup_one_bio(reg, &wc, &current_slot, max_slots,
597 					 REQ_OP_READ, 0);
598 		if (IS_ERR(bio)) {
599 			status = PTR_ERR(bio);
600 			mlog_errno(status);
601 			goto bail_and_wait;
602 		}
603 
604 		atomic_inc(&wc.wc_num_reqs);
605 		submit_bio(bio);
606 	}
607 
608 	status = 0;
609 
610 bail_and_wait:
611 	o2hb_wait_on_io(reg, &wc);
612 	if (wc.wc_error && !status)
613 		status = wc.wc_error;
614 
615 	return status;
616 }
617 
618 static int o2hb_issue_node_write(struct o2hb_region *reg,
619 				 struct o2hb_bio_wait_ctxt *write_wc)
620 {
621 	int status;
622 	unsigned int slot;
623 	struct bio *bio;
624 
625 	o2hb_bio_wait_init(write_wc);
626 
627 	slot = o2nm_this_node();
628 
629 	bio = o2hb_setup_one_bio(reg, write_wc, &slot, slot+1, REQ_OP_WRITE,
630 				 REQ_SYNC);
631 	if (IS_ERR(bio)) {
632 		status = PTR_ERR(bio);
633 		mlog_errno(status);
634 		goto bail;
635 	}
636 
637 	atomic_inc(&write_wc->wc_num_reqs);
638 	submit_bio(bio);
639 
640 	status = 0;
641 bail:
642 	return status;
643 }
644 
645 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
646 				     struct o2hb_disk_heartbeat_block *hb_block)
647 {
648 	__le32 old_cksum;
649 	u32 ret;
650 
651 	/* We want to compute the block crc with a 0 value in the
652 	 * hb_cksum field. Save it off here and replace after the
653 	 * crc. */
654 	old_cksum = hb_block->hb_cksum;
655 	hb_block->hb_cksum = 0;
656 
657 	ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
658 
659 	hb_block->hb_cksum = old_cksum;
660 
661 	return ret;
662 }
663 
664 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
665 {
666 	mlog(ML_ERROR, "Dump slot information: seq = 0x%llx, node = %u, "
667 	     "cksum = 0x%x, generation 0x%llx\n",
668 	     (long long)le64_to_cpu(hb_block->hb_seq),
669 	     hb_block->hb_node, le32_to_cpu(hb_block->hb_cksum),
670 	     (long long)le64_to_cpu(hb_block->hb_generation));
671 }
672 
673 static int o2hb_verify_crc(struct o2hb_region *reg,
674 			   struct o2hb_disk_heartbeat_block *hb_block)
675 {
676 	u32 read, computed;
677 
678 	read = le32_to_cpu(hb_block->hb_cksum);
679 	computed = o2hb_compute_block_crc_le(reg, hb_block);
680 
681 	return read == computed;
682 }
683 
684 /*
685  * Compare the slot data with what we wrote in the last iteration.
686  * If the match fails, print an appropriate error message. This is to
687  * detect errors like... another node hearting on the same slot,
688  * flaky device that is losing writes, etc.
689  * Returns 1 if check succeeds, 0 otherwise.
690  */
691 static int o2hb_check_own_slot(struct o2hb_region *reg)
692 {
693 	struct o2hb_disk_slot *slot;
694 	struct o2hb_disk_heartbeat_block *hb_block;
695 	char *errstr;
696 
697 	slot = &reg->hr_slots[o2nm_this_node()];
698 	/* Don't check on our 1st timestamp */
699 	if (!slot->ds_last_time)
700 		return 0;
701 
702 	hb_block = slot->ds_raw_block;
703 	if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time &&
704 	    le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation &&
705 	    hb_block->hb_node == slot->ds_node_num)
706 		return 1;
707 
708 #define ERRSTR1		"Another node is heartbeating on device"
709 #define ERRSTR2		"Heartbeat generation mismatch on device"
710 #define ERRSTR3		"Heartbeat sequence mismatch on device"
711 
712 	if (hb_block->hb_node != slot->ds_node_num)
713 		errstr = ERRSTR1;
714 	else if (le64_to_cpu(hb_block->hb_generation) !=
715 		 slot->ds_last_generation)
716 		errstr = ERRSTR2;
717 	else
718 		errstr = ERRSTR3;
719 
720 	mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), "
721 	     "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name,
722 	     slot->ds_node_num, (unsigned long long)slot->ds_last_generation,
723 	     (unsigned long long)slot->ds_last_time, hb_block->hb_node,
724 	     (unsigned long long)le64_to_cpu(hb_block->hb_generation),
725 	     (unsigned long long)le64_to_cpu(hb_block->hb_seq));
726 
727 	return 0;
728 }
729 
730 static inline void o2hb_prepare_block(struct o2hb_region *reg,
731 				      u64 generation)
732 {
733 	int node_num;
734 	u64 cputime;
735 	struct o2hb_disk_slot *slot;
736 	struct o2hb_disk_heartbeat_block *hb_block;
737 
738 	node_num = o2nm_this_node();
739 	slot = &reg->hr_slots[node_num];
740 
741 	hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
742 	memset(hb_block, 0, reg->hr_block_bytes);
743 	/* TODO: time stuff */
744 	cputime = ktime_get_real_seconds();
745 	if (!cputime)
746 		cputime = 1;
747 
748 	hb_block->hb_seq = cpu_to_le64(cputime);
749 	hb_block->hb_node = node_num;
750 	hb_block->hb_generation = cpu_to_le64(generation);
751 	hb_block->hb_dead_ms = cpu_to_le32(o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS);
752 
753 	/* This step must always happen last! */
754 	hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
755 								   hb_block));
756 
757 	mlog(ML_HB_BIO, "our node generation = 0x%llx, cksum = 0x%x\n",
758 	     (long long)generation,
759 	     le32_to_cpu(hb_block->hb_cksum));
760 }
761 
762 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
763 				struct o2nm_node *node,
764 				int idx)
765 {
766 	struct o2hb_callback_func *f;
767 
768 	list_for_each_entry(f, &hbcall->list, hc_item) {
769 		mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
770 		(f->hc_func)(node, idx, f->hc_data);
771 	}
772 }
773 
774 /* Will run the list in order until we process the passed event */
775 static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
776 {
777 	struct o2hb_callback *hbcall;
778 	struct o2hb_node_event *event;
779 
780 	/* Holding callback sem assures we don't alter the callback
781 	 * lists when doing this, and serializes ourselves with other
782 	 * processes wanting callbacks. */
783 	down_write(&o2hb_callback_sem);
784 
785 	spin_lock(&o2hb_live_lock);
786 	while (!list_empty(&o2hb_node_events)
787 	       && !list_empty(&queued_event->hn_item)) {
788 		event = list_entry(o2hb_node_events.next,
789 				   struct o2hb_node_event,
790 				   hn_item);
791 		list_del_init(&event->hn_item);
792 		spin_unlock(&o2hb_live_lock);
793 
794 		mlog(ML_HEARTBEAT, "Node %s event for %d\n",
795 		     event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
796 		     event->hn_node_num);
797 
798 		hbcall = hbcall_from_type(event->hn_event_type);
799 
800 		/* We should *never* have gotten on to the list with a
801 		 * bad type... This isn't something that we should try
802 		 * to recover from. */
803 		BUG_ON(IS_ERR(hbcall));
804 
805 		o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
806 
807 		spin_lock(&o2hb_live_lock);
808 	}
809 	spin_unlock(&o2hb_live_lock);
810 
811 	up_write(&o2hb_callback_sem);
812 }
813 
814 static void o2hb_queue_node_event(struct o2hb_node_event *event,
815 				  enum o2hb_callback_type type,
816 				  struct o2nm_node *node,
817 				  int node_num)
818 {
819 	assert_spin_locked(&o2hb_live_lock);
820 
821 	BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB));
822 
823 	event->hn_event_type = type;
824 	event->hn_node = node;
825 	event->hn_node_num = node_num;
826 
827 	mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
828 	     type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
829 
830 	list_add_tail(&event->hn_item, &o2hb_node_events);
831 }
832 
833 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
834 {
835 	struct o2hb_node_event event =
836 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
837 	struct o2nm_node *node;
838 	int queued = 0;
839 
840 	node = o2nm_get_node_by_num(slot->ds_node_num);
841 	if (!node)
842 		return;
843 
844 	spin_lock(&o2hb_live_lock);
845 	if (!list_empty(&slot->ds_live_item)) {
846 		mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
847 		     slot->ds_node_num);
848 
849 		list_del_init(&slot->ds_live_item);
850 
851 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
852 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
853 
854 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
855 					      slot->ds_node_num);
856 			queued = 1;
857 		}
858 	}
859 	spin_unlock(&o2hb_live_lock);
860 
861 	if (queued)
862 		o2hb_run_event_list(&event);
863 
864 	o2nm_node_put(node);
865 }
866 
867 static void o2hb_set_quorum_device(struct o2hb_region *reg)
868 {
869 	if (!o2hb_global_heartbeat_active())
870 		return;
871 
872 	/* Prevent race with o2hb_heartbeat_group_drop_item() */
873 	if (kthread_should_stop())
874 		return;
875 
876 	/* Tag region as quorum only after thread reaches steady state */
877 	if (atomic_read(&reg->hr_steady_iterations) != 0)
878 		return;
879 
880 	spin_lock(&o2hb_live_lock);
881 
882 	if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
883 		goto unlock;
884 
885 	/*
886 	 * A region can be added to the quorum only when it sees all
887 	 * live nodes heartbeat on it. In other words, the region has been
888 	 * added to all nodes.
889 	 */
890 	if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap,
891 		   sizeof(o2hb_live_node_bitmap)))
892 		goto unlock;
893 
894 	printk(KERN_NOTICE "o2hb: Region %s (%s) is now a quorum device\n",
895 	       config_item_name(&reg->hr_item), reg->hr_dev_name);
896 
897 	set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
898 
899 	/*
900 	 * If global heartbeat active, unpin all regions if the
901 	 * region count > CUT_OFF
902 	 */
903 	if (bitmap_weight(o2hb_quorum_region_bitmap,
904 			   O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF)
905 		o2hb_region_unpin(NULL);
906 unlock:
907 	spin_unlock(&o2hb_live_lock);
908 }
909 
910 static int o2hb_check_slot(struct o2hb_region *reg,
911 			   struct o2hb_disk_slot *slot)
912 {
913 	int changed = 0, gen_changed = 0;
914 	struct o2hb_node_event event =
915 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
916 	struct o2nm_node *node;
917 	struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
918 	u64 cputime;
919 	unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS;
920 	unsigned int slot_dead_ms;
921 	int tmp;
922 	int queued = 0;
923 
924 	memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
925 
926 	/*
927 	 * If a node is no longer configured but is still in the livemap, we
928 	 * may need to clear that bit from the livemap.
929 	 */
930 	node = o2nm_get_node_by_num(slot->ds_node_num);
931 	if (!node) {
932 		spin_lock(&o2hb_live_lock);
933 		tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap);
934 		spin_unlock(&o2hb_live_lock);
935 		if (!tmp)
936 			return 0;
937 	}
938 
939 	if (!o2hb_verify_crc(reg, hb_block)) {
940 		/* all paths from here will drop o2hb_live_lock for
941 		 * us. */
942 		spin_lock(&o2hb_live_lock);
943 
944 		/* Don't print an error on the console in this case -
945 		 * a freshly formatted heartbeat area will not have a
946 		 * crc set on it. */
947 		if (list_empty(&slot->ds_live_item))
948 			goto out;
949 
950 		/* The node is live but pushed out a bad crc. We
951 		 * consider it a transient miss but don't populate any
952 		 * other values as they may be junk. */
953 		mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
954 		     slot->ds_node_num, reg->hr_dev_name);
955 		o2hb_dump_slot(hb_block);
956 
957 		slot->ds_equal_samples++;
958 		goto fire_callbacks;
959 	}
960 
961 	/* we don't care if these wrap.. the state transitions below
962 	 * clear at the right places */
963 	cputime = le64_to_cpu(hb_block->hb_seq);
964 	if (slot->ds_last_time != cputime)
965 		slot->ds_changed_samples++;
966 	else
967 		slot->ds_equal_samples++;
968 	slot->ds_last_time = cputime;
969 
970 	/* The node changed heartbeat generations. We assume this to
971 	 * mean it dropped off but came back before we timed out. We
972 	 * want to consider it down for the time being but don't want
973 	 * to lose any changed_samples state we might build up to
974 	 * considering it live again. */
975 	if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
976 		gen_changed = 1;
977 		slot->ds_equal_samples = 0;
978 		mlog(ML_HEARTBEAT, "Node %d changed generation (0x%llx "
979 		     "to 0x%llx)\n", slot->ds_node_num,
980 		     (long long)slot->ds_last_generation,
981 		     (long long)le64_to_cpu(hb_block->hb_generation));
982 	}
983 
984 	slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
985 
986 	mlog(ML_HEARTBEAT, "Slot %d gen 0x%llx cksum 0x%x "
987 	     "seq %llu last %llu changed %u equal %u\n",
988 	     slot->ds_node_num, (long long)slot->ds_last_generation,
989 	     le32_to_cpu(hb_block->hb_cksum),
990 	     (unsigned long long)le64_to_cpu(hb_block->hb_seq),
991 	     (unsigned long long)slot->ds_last_time, slot->ds_changed_samples,
992 	     slot->ds_equal_samples);
993 
994 	spin_lock(&o2hb_live_lock);
995 
996 fire_callbacks:
997 	/* dead nodes only come to life after some number of
998 	 * changes at any time during their dead time */
999 	if (list_empty(&slot->ds_live_item) &&
1000 	    slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
1001 		mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n",
1002 		     slot->ds_node_num, (long long)slot->ds_last_generation);
1003 
1004 		set_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
1005 
1006 		/* first on the list generates a callback */
1007 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
1008 			mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes "
1009 			     "bitmap\n", slot->ds_node_num);
1010 			set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
1011 
1012 			o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
1013 					      slot->ds_node_num);
1014 
1015 			changed = 1;
1016 			queued = 1;
1017 		}
1018 
1019 		list_add_tail(&slot->ds_live_item,
1020 			      &o2hb_live_slots[slot->ds_node_num]);
1021 
1022 		slot->ds_equal_samples = 0;
1023 
1024 		/* We want to be sure that all nodes agree on the
1025 		 * number of milliseconds before a node will be
1026 		 * considered dead. The self-fencing timeout is
1027 		 * computed from this value, and a discrepancy might
1028 		 * result in heartbeat calling a node dead when it
1029 		 * hasn't self-fenced yet. */
1030 		slot_dead_ms = le32_to_cpu(hb_block->hb_dead_ms);
1031 		if (slot_dead_ms && slot_dead_ms != dead_ms) {
1032 			/* TODO: Perhaps we can fail the region here. */
1033 			mlog(ML_ERROR, "Node %d on device %s has a dead count "
1034 			     "of %u ms, but our count is %u ms.\n"
1035 			     "Please double check your configuration values "
1036 			     "for 'O2CB_HEARTBEAT_THRESHOLD'\n",
1037 			     slot->ds_node_num, reg->hr_dev_name, slot_dead_ms,
1038 			     dead_ms);
1039 		}
1040 		goto out;
1041 	}
1042 
1043 	/* if the list is dead, we're done.. */
1044 	if (list_empty(&slot->ds_live_item))
1045 		goto out;
1046 
1047 	/* live nodes only go dead after enough consequtive missed
1048 	 * samples..  reset the missed counter whenever we see
1049 	 * activity */
1050 	if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
1051 		mlog(ML_HEARTBEAT, "Node %d left my region\n",
1052 		     slot->ds_node_num);
1053 
1054 		clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
1055 
1056 		/* last off the live_slot generates a callback */
1057 		list_del_init(&slot->ds_live_item);
1058 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
1059 			mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live "
1060 			     "nodes bitmap\n", slot->ds_node_num);
1061 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
1062 
1063 			/* node can be null */
1064 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB,
1065 					      node, slot->ds_node_num);
1066 
1067 			changed = 1;
1068 			queued = 1;
1069 		}
1070 
1071 		/* We don't clear this because the node is still
1072 		 * actually writing new blocks. */
1073 		if (!gen_changed)
1074 			slot->ds_changed_samples = 0;
1075 		goto out;
1076 	}
1077 	if (slot->ds_changed_samples) {
1078 		slot->ds_changed_samples = 0;
1079 		slot->ds_equal_samples = 0;
1080 	}
1081 out:
1082 	spin_unlock(&o2hb_live_lock);
1083 
1084 	if (queued)
1085 		o2hb_run_event_list(&event);
1086 
1087 	if (node)
1088 		o2nm_node_put(node);
1089 	return changed;
1090 }
1091 
1092 static int o2hb_highest_node(unsigned long *nodes, int numbits)
1093 {
1094 	return find_last_bit(nodes, numbits);
1095 }
1096 
1097 static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
1098 {
1099 	int i, ret, highest_node;
1100 	int membership_change = 0, own_slot_ok = 0;
1101 	unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
1102 	unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
1103 	struct o2hb_bio_wait_ctxt write_wc;
1104 
1105 	ret = o2nm_configured_node_map(configured_nodes,
1106 				       sizeof(configured_nodes));
1107 	if (ret) {
1108 		mlog_errno(ret);
1109 		goto bail;
1110 	}
1111 
1112 	/*
1113 	 * If a node is not configured but is in the livemap, we still need
1114 	 * to read the slot so as to be able to remove it from the livemap.
1115 	 */
1116 	o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
1117 	i = -1;
1118 	while ((i = find_next_bit(live_node_bitmap,
1119 				  O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1120 		set_bit(i, configured_nodes);
1121 	}
1122 
1123 	highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
1124 	if (highest_node >= O2NM_MAX_NODES) {
1125 		mlog(ML_NOTICE, "o2hb: No configured nodes found!\n");
1126 		ret = -EINVAL;
1127 		goto bail;
1128 	}
1129 
1130 	/* No sense in reading the slots of nodes that don't exist
1131 	 * yet. Of course, if the node definitions have holes in them
1132 	 * then we're reading an empty slot anyway... Consider this
1133 	 * best-effort. */
1134 	ret = o2hb_read_slots(reg, highest_node + 1);
1135 	if (ret < 0) {
1136 		mlog_errno(ret);
1137 		goto bail;
1138 	}
1139 
1140 	/* With an up to date view of the slots, we can check that no
1141 	 * other node has been improperly configured to heartbeat in
1142 	 * our slot. */
1143 	own_slot_ok = o2hb_check_own_slot(reg);
1144 
1145 	/* fill in the proper info for our next heartbeat */
1146 	o2hb_prepare_block(reg, reg->hr_generation);
1147 
1148 	ret = o2hb_issue_node_write(reg, &write_wc);
1149 	if (ret < 0) {
1150 		mlog_errno(ret);
1151 		goto bail;
1152 	}
1153 
1154 	i = -1;
1155 	while((i = find_next_bit(configured_nodes,
1156 				 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1157 		membership_change |= o2hb_check_slot(reg, &reg->hr_slots[i]);
1158 	}
1159 
1160 	/*
1161 	 * We have to be sure we've advertised ourselves on disk
1162 	 * before we can go to steady state.  This ensures that
1163 	 * people we find in our steady state have seen us.
1164 	 */
1165 	o2hb_wait_on_io(reg, &write_wc);
1166 	if (write_wc.wc_error) {
1167 		/* Do not re-arm the write timeout on I/O error - we
1168 		 * can't be sure that the new block ever made it to
1169 		 * disk */
1170 		mlog(ML_ERROR, "Write error %d on device \"%s\"\n",
1171 		     write_wc.wc_error, reg->hr_dev_name);
1172 		ret = write_wc.wc_error;
1173 		goto bail;
1174 	}
1175 
1176 	/* Skip disarming the timeout if own slot has stale/bad data */
1177 	if (own_slot_ok) {
1178 		o2hb_set_quorum_device(reg);
1179 		o2hb_arm_timeout(reg);
1180 		reg->hr_last_timeout_start = jiffies;
1181 	}
1182 
1183 bail:
1184 	/* let the person who launched us know when things are steady */
1185 	if (atomic_read(&reg->hr_steady_iterations) != 0) {
1186 		if (!ret && own_slot_ok && !membership_change) {
1187 			if (atomic_dec_and_test(&reg->hr_steady_iterations))
1188 				wake_up(&o2hb_steady_queue);
1189 		}
1190 	}
1191 
1192 	if (atomic_read(&reg->hr_steady_iterations) != 0) {
1193 		if (atomic_dec_and_test(&reg->hr_unsteady_iterations)) {
1194 			printk(KERN_NOTICE "o2hb: Unable to stabilize "
1195 			       "heartbeart on region %s (%s)\n",
1196 			       config_item_name(&reg->hr_item),
1197 			       reg->hr_dev_name);
1198 			atomic_set(&reg->hr_steady_iterations, 0);
1199 			reg->hr_aborted_start = 1;
1200 			wake_up(&o2hb_steady_queue);
1201 			ret = -EIO;
1202 		}
1203 	}
1204 
1205 	return ret;
1206 }
1207 
1208 /*
1209  * we ride the region ref that the region dir holds.  before the region
1210  * dir is removed and drops it ref it will wait to tear down this
1211  * thread.
1212  */
1213 static int o2hb_thread(void *data)
1214 {
1215 	int i, ret;
1216 	struct o2hb_region *reg = data;
1217 	struct o2hb_bio_wait_ctxt write_wc;
1218 	ktime_t before_hb, after_hb;
1219 	unsigned int elapsed_msec;
1220 
1221 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
1222 
1223 	set_user_nice(current, MIN_NICE);
1224 
1225 	/* Pin node */
1226 	ret = o2nm_depend_this_node();
1227 	if (ret) {
1228 		mlog(ML_ERROR, "Node has been deleted, ret = %d\n", ret);
1229 		reg->hr_node_deleted = 1;
1230 		wake_up(&o2hb_steady_queue);
1231 		return 0;
1232 	}
1233 
1234 	while (!kthread_should_stop() &&
1235 	       !reg->hr_unclean_stop && !reg->hr_aborted_start) {
1236 		/* We track the time spent inside
1237 		 * o2hb_do_disk_heartbeat so that we avoid more than
1238 		 * hr_timeout_ms between disk writes. On busy systems
1239 		 * this should result in a heartbeat which is less
1240 		 * likely to time itself out. */
1241 		before_hb = ktime_get_real();
1242 
1243 		ret = o2hb_do_disk_heartbeat(reg);
1244 		reg->hr_last_hb_status = ret;
1245 
1246 		after_hb = ktime_get_real();
1247 
1248 		elapsed_msec = (unsigned int)
1249 				ktime_ms_delta(after_hb, before_hb);
1250 
1251 		mlog(ML_HEARTBEAT,
1252 		     "start = %lld, end = %lld, msec = %u, ret = %d\n",
1253 		     before_hb, after_hb, elapsed_msec, ret);
1254 
1255 		if (!kthread_should_stop() &&
1256 		    elapsed_msec < reg->hr_timeout_ms) {
1257 			/* the kthread api has blocked signals for us so no
1258 			 * need to record the return value. */
1259 			msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
1260 		}
1261 	}
1262 
1263 	o2hb_disarm_timeout(reg);
1264 
1265 	/* unclean stop is only used in very bad situation */
1266 	for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
1267 		o2hb_shutdown_slot(&reg->hr_slots[i]);
1268 
1269 	/* Explicit down notification - avoid forcing the other nodes
1270 	 * to timeout on this region when we could just as easily
1271 	 * write a clear generation - thus indicating to them that
1272 	 * this node has left this region.
1273 	 */
1274 	if (!reg->hr_unclean_stop && !reg->hr_aborted_start) {
1275 		o2hb_prepare_block(reg, 0);
1276 		ret = o2hb_issue_node_write(reg, &write_wc);
1277 		if (ret == 0)
1278 			o2hb_wait_on_io(reg, &write_wc);
1279 		else
1280 			mlog_errno(ret);
1281 	}
1282 
1283 	/* Unpin node */
1284 	o2nm_undepend_this_node();
1285 
1286 	mlog(ML_HEARTBEAT|ML_KTHREAD, "o2hb thread exiting\n");
1287 
1288 	return 0;
1289 }
1290 
1291 #ifdef CONFIG_DEBUG_FS
1292 static int o2hb_debug_open(struct inode *inode, struct file *file)
1293 {
1294 	struct o2hb_debug_buf *db = inode->i_private;
1295 	struct o2hb_region *reg;
1296 	unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1297 	unsigned long lts;
1298 	char *buf = NULL;
1299 	int i = -1;
1300 	int out = 0;
1301 
1302 	/* max_nodes should be the largest bitmap we pass here */
1303 	BUG_ON(sizeof(map) < db->db_size);
1304 
1305 	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
1306 	if (!buf)
1307 		goto bail;
1308 
1309 	switch (db->db_type) {
1310 	case O2HB_DB_TYPE_LIVENODES:
1311 	case O2HB_DB_TYPE_LIVEREGIONS:
1312 	case O2HB_DB_TYPE_QUORUMREGIONS:
1313 	case O2HB_DB_TYPE_FAILEDREGIONS:
1314 		spin_lock(&o2hb_live_lock);
1315 		memcpy(map, db->db_data, db->db_size);
1316 		spin_unlock(&o2hb_live_lock);
1317 		break;
1318 
1319 	case O2HB_DB_TYPE_REGION_LIVENODES:
1320 		spin_lock(&o2hb_live_lock);
1321 		reg = (struct o2hb_region *)db->db_data;
1322 		memcpy(map, reg->hr_live_node_bitmap, db->db_size);
1323 		spin_unlock(&o2hb_live_lock);
1324 		break;
1325 
1326 	case O2HB_DB_TYPE_REGION_NUMBER:
1327 		reg = (struct o2hb_region *)db->db_data;
1328 		out += snprintf(buf + out, PAGE_SIZE - out, "%d\n",
1329 				reg->hr_region_num);
1330 		goto done;
1331 
1332 	case O2HB_DB_TYPE_REGION_ELAPSED_TIME:
1333 		reg = (struct o2hb_region *)db->db_data;
1334 		lts = reg->hr_last_timeout_start;
1335 		/* If 0, it has never been set before */
1336 		if (lts)
1337 			lts = jiffies_to_msecs(jiffies - lts);
1338 		out += snprintf(buf + out, PAGE_SIZE - out, "%lu\n", lts);
1339 		goto done;
1340 
1341 	case O2HB_DB_TYPE_REGION_PINNED:
1342 		reg = (struct o2hb_region *)db->db_data;
1343 		out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
1344 				!!reg->hr_item_pinned);
1345 		goto done;
1346 
1347 	default:
1348 		goto done;
1349 	}
1350 
1351 	while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len)
1352 		out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i);
1353 	out += snprintf(buf + out, PAGE_SIZE - out, "\n");
1354 
1355 done:
1356 	i_size_write(inode, out);
1357 
1358 	file->private_data = buf;
1359 
1360 	return 0;
1361 bail:
1362 	return -ENOMEM;
1363 }
1364 
1365 static int o2hb_debug_release(struct inode *inode, struct file *file)
1366 {
1367 	kfree(file->private_data);
1368 	return 0;
1369 }
1370 
1371 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1372 				 size_t nbytes, loff_t *ppos)
1373 {
1374 	return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
1375 				       i_size_read(file->f_mapping->host));
1376 }
1377 #else
1378 static int o2hb_debug_open(struct inode *inode, struct file *file)
1379 {
1380 	return 0;
1381 }
1382 static int o2hb_debug_release(struct inode *inode, struct file *file)
1383 {
1384 	return 0;
1385 }
1386 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1387 			       size_t nbytes, loff_t *ppos)
1388 {
1389 	return 0;
1390 }
1391 #endif  /* CONFIG_DEBUG_FS */
1392 
1393 static const struct file_operations o2hb_debug_fops = {
1394 	.open =		o2hb_debug_open,
1395 	.release =	o2hb_debug_release,
1396 	.read =		o2hb_debug_read,
1397 	.llseek =	generic_file_llseek,
1398 };
1399 
1400 void o2hb_exit(void)
1401 {
1402 	debugfs_remove(o2hb_debug_failedregions);
1403 	debugfs_remove(o2hb_debug_quorumregions);
1404 	debugfs_remove(o2hb_debug_liveregions);
1405 	debugfs_remove(o2hb_debug_livenodes);
1406 	debugfs_remove(o2hb_debug_dir);
1407 	kfree(o2hb_db_livenodes);
1408 	kfree(o2hb_db_liveregions);
1409 	kfree(o2hb_db_quorumregions);
1410 	kfree(o2hb_db_failedregions);
1411 }
1412 
1413 static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir,
1414 					struct o2hb_debug_buf **db, int db_len,
1415 					int type, int size, int len, void *data)
1416 {
1417 	*db = kmalloc(db_len, GFP_KERNEL);
1418 	if (!*db)
1419 		return NULL;
1420 
1421 	(*db)->db_type = type;
1422 	(*db)->db_size = size;
1423 	(*db)->db_len = len;
1424 	(*db)->db_data = data;
1425 
1426 	return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db,
1427 				   &o2hb_debug_fops);
1428 }
1429 
1430 static int o2hb_debug_init(void)
1431 {
1432 	int ret = -ENOMEM;
1433 
1434 	o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL);
1435 	if (!o2hb_debug_dir) {
1436 		mlog_errno(ret);
1437 		goto bail;
1438 	}
1439 
1440 	o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES,
1441 						 o2hb_debug_dir,
1442 						 &o2hb_db_livenodes,
1443 						 sizeof(*o2hb_db_livenodes),
1444 						 O2HB_DB_TYPE_LIVENODES,
1445 						 sizeof(o2hb_live_node_bitmap),
1446 						 O2NM_MAX_NODES,
1447 						 o2hb_live_node_bitmap);
1448 	if (!o2hb_debug_livenodes) {
1449 		mlog_errno(ret);
1450 		goto bail;
1451 	}
1452 
1453 	o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS,
1454 						   o2hb_debug_dir,
1455 						   &o2hb_db_liveregions,
1456 						   sizeof(*o2hb_db_liveregions),
1457 						   O2HB_DB_TYPE_LIVEREGIONS,
1458 						   sizeof(o2hb_live_region_bitmap),
1459 						   O2NM_MAX_REGIONS,
1460 						   o2hb_live_region_bitmap);
1461 	if (!o2hb_debug_liveregions) {
1462 		mlog_errno(ret);
1463 		goto bail;
1464 	}
1465 
1466 	o2hb_debug_quorumregions =
1467 			o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS,
1468 					  o2hb_debug_dir,
1469 					  &o2hb_db_quorumregions,
1470 					  sizeof(*o2hb_db_quorumregions),
1471 					  O2HB_DB_TYPE_QUORUMREGIONS,
1472 					  sizeof(o2hb_quorum_region_bitmap),
1473 					  O2NM_MAX_REGIONS,
1474 					  o2hb_quorum_region_bitmap);
1475 	if (!o2hb_debug_quorumregions) {
1476 		mlog_errno(ret);
1477 		goto bail;
1478 	}
1479 
1480 	o2hb_debug_failedregions =
1481 			o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS,
1482 					  o2hb_debug_dir,
1483 					  &o2hb_db_failedregions,
1484 					  sizeof(*o2hb_db_failedregions),
1485 					  O2HB_DB_TYPE_FAILEDREGIONS,
1486 					  sizeof(o2hb_failed_region_bitmap),
1487 					  O2NM_MAX_REGIONS,
1488 					  o2hb_failed_region_bitmap);
1489 	if (!o2hb_debug_failedregions) {
1490 		mlog_errno(ret);
1491 		goto bail;
1492 	}
1493 
1494 	ret = 0;
1495 bail:
1496 	if (ret)
1497 		o2hb_exit();
1498 
1499 	return ret;
1500 }
1501 
1502 int o2hb_init(void)
1503 {
1504 	int i;
1505 
1506 	for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
1507 		INIT_LIST_HEAD(&o2hb_callbacks[i].list);
1508 
1509 	for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
1510 		INIT_LIST_HEAD(&o2hb_live_slots[i]);
1511 
1512 	INIT_LIST_HEAD(&o2hb_node_events);
1513 
1514 	memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
1515 	memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap));
1516 	memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap));
1517 	memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap));
1518 	memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap));
1519 
1520 	o2hb_dependent_users = 0;
1521 
1522 	return o2hb_debug_init();
1523 }
1524 
1525 /* if we're already in a callback then we're already serialized by the sem */
1526 static void o2hb_fill_node_map_from_callback(unsigned long *map,
1527 					     unsigned bytes)
1528 {
1529 	BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
1530 
1531 	memcpy(map, &o2hb_live_node_bitmap, bytes);
1532 }
1533 
1534 /*
1535  * get a map of all nodes that are heartbeating in any regions
1536  */
1537 void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
1538 {
1539 	/* callers want to serialize this map and callbacks so that they
1540 	 * can trust that they don't miss nodes coming to the party */
1541 	down_read(&o2hb_callback_sem);
1542 	spin_lock(&o2hb_live_lock);
1543 	o2hb_fill_node_map_from_callback(map, bytes);
1544 	spin_unlock(&o2hb_live_lock);
1545 	up_read(&o2hb_callback_sem);
1546 }
1547 EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
1548 
1549 /*
1550  * heartbeat configfs bits.  The heartbeat set is a default set under
1551  * the cluster set in nodemanager.c.
1552  */
1553 
1554 static struct o2hb_region *to_o2hb_region(struct config_item *item)
1555 {
1556 	return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
1557 }
1558 
1559 /* drop_item only drops its ref after killing the thread, nothing should
1560  * be using the region anymore.  this has to clean up any state that
1561  * attributes might have built up. */
1562 static void o2hb_region_release(struct config_item *item)
1563 {
1564 	int i;
1565 	struct page *page;
1566 	struct o2hb_region *reg = to_o2hb_region(item);
1567 
1568 	mlog(ML_HEARTBEAT, "hb region release (%s)\n", reg->hr_dev_name);
1569 
1570 	kfree(reg->hr_tmp_block);
1571 
1572 	if (reg->hr_slot_data) {
1573 		for (i = 0; i < reg->hr_num_pages; i++) {
1574 			page = reg->hr_slot_data[i];
1575 			if (page)
1576 				__free_page(page);
1577 		}
1578 		kfree(reg->hr_slot_data);
1579 	}
1580 
1581 	if (reg->hr_bdev)
1582 		blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1583 
1584 	kfree(reg->hr_slots);
1585 
1586 	debugfs_remove(reg->hr_debug_livenodes);
1587 	debugfs_remove(reg->hr_debug_regnum);
1588 	debugfs_remove(reg->hr_debug_elapsed_time);
1589 	debugfs_remove(reg->hr_debug_pinned);
1590 	debugfs_remove(reg->hr_debug_dir);
1591 	kfree(reg->hr_db_livenodes);
1592 	kfree(reg->hr_db_regnum);
1593 	kfree(reg->hr_db_elapsed_time);
1594 	kfree(reg->hr_db_pinned);
1595 
1596 	spin_lock(&o2hb_live_lock);
1597 	list_del(&reg->hr_all_item);
1598 	spin_unlock(&o2hb_live_lock);
1599 
1600 	o2net_unregister_handler_list(&reg->hr_handler_list);
1601 	kfree(reg);
1602 }
1603 
1604 static int o2hb_read_block_input(struct o2hb_region *reg,
1605 				 const char *page,
1606 				 unsigned long *ret_bytes,
1607 				 unsigned int *ret_bits)
1608 {
1609 	unsigned long bytes;
1610 	char *p = (char *)page;
1611 
1612 	bytes = simple_strtoul(p, &p, 0);
1613 	if (!p || (*p && (*p != '\n')))
1614 		return -EINVAL;
1615 
1616 	/* Heartbeat and fs min / max block sizes are the same. */
1617 	if (bytes > 4096 || bytes < 512)
1618 		return -ERANGE;
1619 	if (hweight16(bytes) != 1)
1620 		return -EINVAL;
1621 
1622 	if (ret_bytes)
1623 		*ret_bytes = bytes;
1624 	if (ret_bits)
1625 		*ret_bits = ffs(bytes) - 1;
1626 
1627 	return 0;
1628 }
1629 
1630 static ssize_t o2hb_region_block_bytes_show(struct config_item *item,
1631 					    char *page)
1632 {
1633 	return sprintf(page, "%u\n", to_o2hb_region(item)->hr_block_bytes);
1634 }
1635 
1636 static ssize_t o2hb_region_block_bytes_store(struct config_item *item,
1637 					     const char *page,
1638 					     size_t count)
1639 {
1640 	struct o2hb_region *reg = to_o2hb_region(item);
1641 	int status;
1642 	unsigned long block_bytes;
1643 	unsigned int block_bits;
1644 
1645 	if (reg->hr_bdev)
1646 		return -EINVAL;
1647 
1648 	status = o2hb_read_block_input(reg, page, &block_bytes,
1649 				       &block_bits);
1650 	if (status)
1651 		return status;
1652 
1653 	reg->hr_block_bytes = (unsigned int)block_bytes;
1654 	reg->hr_block_bits = block_bits;
1655 
1656 	return count;
1657 }
1658 
1659 static ssize_t o2hb_region_start_block_show(struct config_item *item,
1660 					    char *page)
1661 {
1662 	return sprintf(page, "%llu\n", to_o2hb_region(item)->hr_start_block);
1663 }
1664 
1665 static ssize_t o2hb_region_start_block_store(struct config_item *item,
1666 					     const char *page,
1667 					     size_t count)
1668 {
1669 	struct o2hb_region *reg = to_o2hb_region(item);
1670 	unsigned long long tmp;
1671 	char *p = (char *)page;
1672 
1673 	if (reg->hr_bdev)
1674 		return -EINVAL;
1675 
1676 	tmp = simple_strtoull(p, &p, 0);
1677 	if (!p || (*p && (*p != '\n')))
1678 		return -EINVAL;
1679 
1680 	reg->hr_start_block = tmp;
1681 
1682 	return count;
1683 }
1684 
1685 static ssize_t o2hb_region_blocks_show(struct config_item *item, char *page)
1686 {
1687 	return sprintf(page, "%d\n", to_o2hb_region(item)->hr_blocks);
1688 }
1689 
1690 static ssize_t o2hb_region_blocks_store(struct config_item *item,
1691 					const char *page,
1692 					size_t count)
1693 {
1694 	struct o2hb_region *reg = to_o2hb_region(item);
1695 	unsigned long tmp;
1696 	char *p = (char *)page;
1697 
1698 	if (reg->hr_bdev)
1699 		return -EINVAL;
1700 
1701 	tmp = simple_strtoul(p, &p, 0);
1702 	if (!p || (*p && (*p != '\n')))
1703 		return -EINVAL;
1704 
1705 	if (tmp > O2NM_MAX_NODES || tmp == 0)
1706 		return -ERANGE;
1707 
1708 	reg->hr_blocks = (unsigned int)tmp;
1709 
1710 	return count;
1711 }
1712 
1713 static ssize_t o2hb_region_dev_show(struct config_item *item, char *page)
1714 {
1715 	unsigned int ret = 0;
1716 
1717 	if (to_o2hb_region(item)->hr_bdev)
1718 		ret = sprintf(page, "%s\n", to_o2hb_region(item)->hr_dev_name);
1719 
1720 	return ret;
1721 }
1722 
1723 static void o2hb_init_region_params(struct o2hb_region *reg)
1724 {
1725 	reg->hr_slots_per_page = PAGE_SIZE >> reg->hr_block_bits;
1726 	reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
1727 
1728 	mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
1729 	     reg->hr_start_block, reg->hr_blocks);
1730 	mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
1731 	     reg->hr_block_bytes, reg->hr_block_bits);
1732 	mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
1733 	mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
1734 }
1735 
1736 static int o2hb_map_slot_data(struct o2hb_region *reg)
1737 {
1738 	int i, j;
1739 	unsigned int last_slot;
1740 	unsigned int spp = reg->hr_slots_per_page;
1741 	struct page *page;
1742 	char *raw;
1743 	struct o2hb_disk_slot *slot;
1744 
1745 	reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
1746 	if (reg->hr_tmp_block == NULL)
1747 		return -ENOMEM;
1748 
1749 	reg->hr_slots = kcalloc(reg->hr_blocks,
1750 				sizeof(struct o2hb_disk_slot), GFP_KERNEL);
1751 	if (reg->hr_slots == NULL)
1752 		return -ENOMEM;
1753 
1754 	for(i = 0; i < reg->hr_blocks; i++) {
1755 		slot = &reg->hr_slots[i];
1756 		slot->ds_node_num = i;
1757 		INIT_LIST_HEAD(&slot->ds_live_item);
1758 		slot->ds_raw_block = NULL;
1759 	}
1760 
1761 	reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
1762 	mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
1763 			   "at %u blocks per page\n",
1764 	     reg->hr_num_pages, reg->hr_blocks, spp);
1765 
1766 	reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
1767 				    GFP_KERNEL);
1768 	if (!reg->hr_slot_data)
1769 		return -ENOMEM;
1770 
1771 	for(i = 0; i < reg->hr_num_pages; i++) {
1772 		page = alloc_page(GFP_KERNEL);
1773 		if (!page)
1774 			return -ENOMEM;
1775 
1776 		reg->hr_slot_data[i] = page;
1777 
1778 		last_slot = i * spp;
1779 		raw = page_address(page);
1780 		for (j = 0;
1781 		     (j < spp) && ((j + last_slot) < reg->hr_blocks);
1782 		     j++) {
1783 			BUG_ON((j + last_slot) >= reg->hr_blocks);
1784 
1785 			slot = &reg->hr_slots[j + last_slot];
1786 			slot->ds_raw_block =
1787 				(struct o2hb_disk_heartbeat_block *) raw;
1788 
1789 			raw += reg->hr_block_bytes;
1790 		}
1791 	}
1792 
1793 	return 0;
1794 }
1795 
1796 /* Read in all the slots available and populate the tracking
1797  * structures so that we can start with a baseline idea of what's
1798  * there. */
1799 static int o2hb_populate_slot_data(struct o2hb_region *reg)
1800 {
1801 	int ret, i;
1802 	struct o2hb_disk_slot *slot;
1803 	struct o2hb_disk_heartbeat_block *hb_block;
1804 
1805 	ret = o2hb_read_slots(reg, reg->hr_blocks);
1806 	if (ret)
1807 		goto out;
1808 
1809 	/* We only want to get an idea of the values initially in each
1810 	 * slot, so we do no verification - o2hb_check_slot will
1811 	 * actually determine if each configured slot is valid and
1812 	 * whether any values have changed. */
1813 	for(i = 0; i < reg->hr_blocks; i++) {
1814 		slot = &reg->hr_slots[i];
1815 		hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
1816 
1817 		/* Only fill the values that o2hb_check_slot uses to
1818 		 * determine changing slots */
1819 		slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
1820 		slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
1821 	}
1822 
1823 out:
1824 	return ret;
1825 }
1826 
1827 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
1828 static ssize_t o2hb_region_dev_store(struct config_item *item,
1829 				     const char *page,
1830 				     size_t count)
1831 {
1832 	struct o2hb_region *reg = to_o2hb_region(item);
1833 	struct task_struct *hb_task;
1834 	long fd;
1835 	int sectsize;
1836 	char *p = (char *)page;
1837 	struct fd f;
1838 	struct inode *inode;
1839 	ssize_t ret = -EINVAL;
1840 	int live_threshold;
1841 
1842 	if (reg->hr_bdev)
1843 		goto out;
1844 
1845 	/* We can't heartbeat without having had our node number
1846 	 * configured yet. */
1847 	if (o2nm_this_node() == O2NM_MAX_NODES)
1848 		goto out;
1849 
1850 	fd = simple_strtol(p, &p, 0);
1851 	if (!p || (*p && (*p != '\n')))
1852 		goto out;
1853 
1854 	if (fd < 0 || fd >= INT_MAX)
1855 		goto out;
1856 
1857 	f = fdget(fd);
1858 	if (f.file == NULL)
1859 		goto out;
1860 
1861 	if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
1862 	    reg->hr_block_bytes == 0)
1863 		goto out2;
1864 
1865 	inode = igrab(f.file->f_mapping->host);
1866 	if (inode == NULL)
1867 		goto out2;
1868 
1869 	if (!S_ISBLK(inode->i_mode))
1870 		goto out3;
1871 
1872 	reg->hr_bdev = I_BDEV(f.file->f_mapping->host);
1873 	ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL);
1874 	if (ret) {
1875 		reg->hr_bdev = NULL;
1876 		goto out3;
1877 	}
1878 	inode = NULL;
1879 
1880 	bdevname(reg->hr_bdev, reg->hr_dev_name);
1881 
1882 	sectsize = bdev_logical_block_size(reg->hr_bdev);
1883 	if (sectsize != reg->hr_block_bytes) {
1884 		mlog(ML_ERROR,
1885 		     "blocksize %u incorrect for device, expected %d",
1886 		     reg->hr_block_bytes, sectsize);
1887 		ret = -EINVAL;
1888 		goto out3;
1889 	}
1890 
1891 	o2hb_init_region_params(reg);
1892 
1893 	/* Generation of zero is invalid */
1894 	do {
1895 		get_random_bytes(&reg->hr_generation,
1896 				 sizeof(reg->hr_generation));
1897 	} while (reg->hr_generation == 0);
1898 
1899 	ret = o2hb_map_slot_data(reg);
1900 	if (ret) {
1901 		mlog_errno(ret);
1902 		goto out3;
1903 	}
1904 
1905 	ret = o2hb_populate_slot_data(reg);
1906 	if (ret) {
1907 		mlog_errno(ret);
1908 		goto out3;
1909 	}
1910 
1911 	INIT_DELAYED_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout);
1912 	INIT_DELAYED_WORK(&reg->hr_nego_timeout_work, o2hb_nego_timeout);
1913 
1914 	/*
1915 	 * A node is considered live after it has beat LIVE_THRESHOLD
1916 	 * times.  We're not steady until we've given them a chance
1917 	 * _after_ our first read.
1918 	 * The default threshold is bare minimum so as to limit the delay
1919 	 * during mounts. For global heartbeat, the threshold doubled for the
1920 	 * first region.
1921 	 */
1922 	live_threshold = O2HB_LIVE_THRESHOLD;
1923 	if (o2hb_global_heartbeat_active()) {
1924 		spin_lock(&o2hb_live_lock);
1925 		if (bitmap_weight(o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1)
1926 			live_threshold <<= 1;
1927 		spin_unlock(&o2hb_live_lock);
1928 	}
1929 	++live_threshold;
1930 	atomic_set(&reg->hr_steady_iterations, live_threshold);
1931 	/* unsteady_iterations is triple the steady_iterations */
1932 	atomic_set(&reg->hr_unsteady_iterations, (live_threshold * 3));
1933 
1934 	hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
1935 			      reg->hr_item.ci_name);
1936 	if (IS_ERR(hb_task)) {
1937 		ret = PTR_ERR(hb_task);
1938 		mlog_errno(ret);
1939 		goto out3;
1940 	}
1941 
1942 	spin_lock(&o2hb_live_lock);
1943 	reg->hr_task = hb_task;
1944 	spin_unlock(&o2hb_live_lock);
1945 
1946 	ret = wait_event_interruptible(o2hb_steady_queue,
1947 				atomic_read(&reg->hr_steady_iterations) == 0 ||
1948 				reg->hr_node_deleted);
1949 	if (ret) {
1950 		atomic_set(&reg->hr_steady_iterations, 0);
1951 		reg->hr_aborted_start = 1;
1952 	}
1953 
1954 	if (reg->hr_aborted_start) {
1955 		ret = -EIO;
1956 		goto out3;
1957 	}
1958 
1959 	if (reg->hr_node_deleted) {
1960 		ret = -EINVAL;
1961 		goto out3;
1962 	}
1963 
1964 	/* Ok, we were woken.  Make sure it wasn't by drop_item() */
1965 	spin_lock(&o2hb_live_lock);
1966 	hb_task = reg->hr_task;
1967 	if (o2hb_global_heartbeat_active())
1968 		set_bit(reg->hr_region_num, o2hb_live_region_bitmap);
1969 	spin_unlock(&o2hb_live_lock);
1970 
1971 	if (hb_task)
1972 		ret = count;
1973 	else
1974 		ret = -EIO;
1975 
1976 	if (hb_task && o2hb_global_heartbeat_active())
1977 		printk(KERN_NOTICE "o2hb: Heartbeat started on region %s (%s)\n",
1978 		       config_item_name(&reg->hr_item), reg->hr_dev_name);
1979 
1980 out3:
1981 	iput(inode);
1982 out2:
1983 	fdput(f);
1984 out:
1985 	if (ret < 0) {
1986 		if (reg->hr_bdev) {
1987 			blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1988 			reg->hr_bdev = NULL;
1989 		}
1990 	}
1991 	return ret;
1992 }
1993 
1994 static ssize_t o2hb_region_pid_show(struct config_item *item, char *page)
1995 {
1996 	struct o2hb_region *reg = to_o2hb_region(item);
1997 	pid_t pid = 0;
1998 
1999 	spin_lock(&o2hb_live_lock);
2000 	if (reg->hr_task)
2001 		pid = task_pid_nr(reg->hr_task);
2002 	spin_unlock(&o2hb_live_lock);
2003 
2004 	if (!pid)
2005 		return 0;
2006 
2007 	return sprintf(page, "%u\n", pid);
2008 }
2009 
2010 CONFIGFS_ATTR(o2hb_region_, block_bytes);
2011 CONFIGFS_ATTR(o2hb_region_, start_block);
2012 CONFIGFS_ATTR(o2hb_region_, blocks);
2013 CONFIGFS_ATTR(o2hb_region_, dev);
2014 CONFIGFS_ATTR_RO(o2hb_region_, pid);
2015 
2016 static struct configfs_attribute *o2hb_region_attrs[] = {
2017 	&o2hb_region_attr_block_bytes,
2018 	&o2hb_region_attr_start_block,
2019 	&o2hb_region_attr_blocks,
2020 	&o2hb_region_attr_dev,
2021 	&o2hb_region_attr_pid,
2022 	NULL,
2023 };
2024 
2025 static struct configfs_item_operations o2hb_region_item_ops = {
2026 	.release		= o2hb_region_release,
2027 };
2028 
2029 static struct config_item_type o2hb_region_type = {
2030 	.ct_item_ops	= &o2hb_region_item_ops,
2031 	.ct_attrs	= o2hb_region_attrs,
2032 	.ct_owner	= THIS_MODULE,
2033 };
2034 
2035 /* heartbeat set */
2036 
2037 struct o2hb_heartbeat_group {
2038 	struct config_group hs_group;
2039 	/* some stuff? */
2040 };
2041 
2042 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
2043 {
2044 	return group ?
2045 		container_of(group, struct o2hb_heartbeat_group, hs_group)
2046 		: NULL;
2047 }
2048 
2049 static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir)
2050 {
2051 	int ret = -ENOMEM;
2052 
2053 	reg->hr_debug_dir =
2054 		debugfs_create_dir(config_item_name(&reg->hr_item), dir);
2055 	if (!reg->hr_debug_dir) {
2056 		mlog_errno(ret);
2057 		goto bail;
2058 	}
2059 
2060 	reg->hr_debug_livenodes =
2061 			o2hb_debug_create(O2HB_DEBUG_LIVENODES,
2062 					  reg->hr_debug_dir,
2063 					  &(reg->hr_db_livenodes),
2064 					  sizeof(*(reg->hr_db_livenodes)),
2065 					  O2HB_DB_TYPE_REGION_LIVENODES,
2066 					  sizeof(reg->hr_live_node_bitmap),
2067 					  O2NM_MAX_NODES, reg);
2068 	if (!reg->hr_debug_livenodes) {
2069 		mlog_errno(ret);
2070 		goto bail;
2071 	}
2072 
2073 	reg->hr_debug_regnum =
2074 			o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER,
2075 					  reg->hr_debug_dir,
2076 					  &(reg->hr_db_regnum),
2077 					  sizeof(*(reg->hr_db_regnum)),
2078 					  O2HB_DB_TYPE_REGION_NUMBER,
2079 					  0, O2NM_MAX_NODES, reg);
2080 	if (!reg->hr_debug_regnum) {
2081 		mlog_errno(ret);
2082 		goto bail;
2083 	}
2084 
2085 	reg->hr_debug_elapsed_time =
2086 			o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME,
2087 					  reg->hr_debug_dir,
2088 					  &(reg->hr_db_elapsed_time),
2089 					  sizeof(*(reg->hr_db_elapsed_time)),
2090 					  O2HB_DB_TYPE_REGION_ELAPSED_TIME,
2091 					  0, 0, reg);
2092 	if (!reg->hr_debug_elapsed_time) {
2093 		mlog_errno(ret);
2094 		goto bail;
2095 	}
2096 
2097 	reg->hr_debug_pinned =
2098 			o2hb_debug_create(O2HB_DEBUG_REGION_PINNED,
2099 					  reg->hr_debug_dir,
2100 					  &(reg->hr_db_pinned),
2101 					  sizeof(*(reg->hr_db_pinned)),
2102 					  O2HB_DB_TYPE_REGION_PINNED,
2103 					  0, 0, reg);
2104 	if (!reg->hr_debug_pinned) {
2105 		mlog_errno(ret);
2106 		goto bail;
2107 	}
2108 
2109 	ret = 0;
2110 bail:
2111 	return ret;
2112 }
2113 
2114 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
2115 							  const char *name)
2116 {
2117 	struct o2hb_region *reg = NULL;
2118 	int ret;
2119 
2120 	reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
2121 	if (reg == NULL)
2122 		return ERR_PTR(-ENOMEM);
2123 
2124 	if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) {
2125 		ret = -ENAMETOOLONG;
2126 		goto free;
2127 	}
2128 
2129 	spin_lock(&o2hb_live_lock);
2130 	reg->hr_region_num = 0;
2131 	if (o2hb_global_heartbeat_active()) {
2132 		reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap,
2133 							 O2NM_MAX_REGIONS);
2134 		if (reg->hr_region_num >= O2NM_MAX_REGIONS) {
2135 			spin_unlock(&o2hb_live_lock);
2136 			ret = -EFBIG;
2137 			goto free;
2138 		}
2139 		set_bit(reg->hr_region_num, o2hb_region_bitmap);
2140 	}
2141 	list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
2142 	spin_unlock(&o2hb_live_lock);
2143 
2144 	config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
2145 
2146 	/* this is the same way to generate msg key as dlm, for local heartbeat,
2147 	 * name is also the same, so make initial crc value different to avoid
2148 	 * message key conflict.
2149 	 */
2150 	reg->hr_key = crc32_le(reg->hr_region_num + O2NM_MAX_REGIONS,
2151 		name, strlen(name));
2152 	INIT_LIST_HEAD(&reg->hr_handler_list);
2153 	ret = o2net_register_handler(O2HB_NEGO_TIMEOUT_MSG, reg->hr_key,
2154 			sizeof(struct o2hb_nego_msg),
2155 			o2hb_nego_timeout_handler,
2156 			reg, NULL, &reg->hr_handler_list);
2157 	if (ret)
2158 		goto free;
2159 
2160 	ret = o2net_register_handler(O2HB_NEGO_APPROVE_MSG, reg->hr_key,
2161 			sizeof(struct o2hb_nego_msg),
2162 			o2hb_nego_approve_handler,
2163 			reg, NULL, &reg->hr_handler_list);
2164 	if (ret)
2165 		goto unregister_handler;
2166 
2167 	ret = o2hb_debug_region_init(reg, o2hb_debug_dir);
2168 	if (ret) {
2169 		config_item_put(&reg->hr_item);
2170 		goto unregister_handler;
2171 	}
2172 
2173 	return &reg->hr_item;
2174 
2175 unregister_handler:
2176 	o2net_unregister_handler_list(&reg->hr_handler_list);
2177 free:
2178 	kfree(reg);
2179 	return ERR_PTR(ret);
2180 }
2181 
2182 static void o2hb_heartbeat_group_drop_item(struct config_group *group,
2183 					   struct config_item *item)
2184 {
2185 	struct task_struct *hb_task;
2186 	struct o2hb_region *reg = to_o2hb_region(item);
2187 	int quorum_region = 0;
2188 
2189 	/* stop the thread when the user removes the region dir */
2190 	spin_lock(&o2hb_live_lock);
2191 	hb_task = reg->hr_task;
2192 	reg->hr_task = NULL;
2193 	reg->hr_item_dropped = 1;
2194 	spin_unlock(&o2hb_live_lock);
2195 
2196 	if (hb_task)
2197 		kthread_stop(hb_task);
2198 
2199 	if (o2hb_global_heartbeat_active()) {
2200 		spin_lock(&o2hb_live_lock);
2201 		clear_bit(reg->hr_region_num, o2hb_region_bitmap);
2202 		clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
2203 		if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
2204 			quorum_region = 1;
2205 		clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
2206 		spin_unlock(&o2hb_live_lock);
2207 		printk(KERN_NOTICE "o2hb: Heartbeat %s on region %s (%s)\n",
2208 		       ((atomic_read(&reg->hr_steady_iterations) == 0) ?
2209 			"stopped" : "start aborted"), config_item_name(item),
2210 		       reg->hr_dev_name);
2211 	}
2212 
2213 	/*
2214 	 * If we're racing a dev_write(), we need to wake them.  They will
2215 	 * check reg->hr_task
2216 	 */
2217 	if (atomic_read(&reg->hr_steady_iterations) != 0) {
2218 		reg->hr_aborted_start = 1;
2219 		atomic_set(&reg->hr_steady_iterations, 0);
2220 		wake_up(&o2hb_steady_queue);
2221 	}
2222 
2223 	config_item_put(item);
2224 
2225 	if (!o2hb_global_heartbeat_active() || !quorum_region)
2226 		return;
2227 
2228 	/*
2229 	 * If global heartbeat active and there are dependent users,
2230 	 * pin all regions if quorum region count <= CUT_OFF
2231 	 */
2232 	spin_lock(&o2hb_live_lock);
2233 
2234 	if (!o2hb_dependent_users)
2235 		goto unlock;
2236 
2237 	if (bitmap_weight(o2hb_quorum_region_bitmap,
2238 			   O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2239 		o2hb_region_pin(NULL);
2240 
2241 unlock:
2242 	spin_unlock(&o2hb_live_lock);
2243 }
2244 
2245 static ssize_t o2hb_heartbeat_group_dead_threshold_show(struct config_item *item,
2246 		char *page)
2247 {
2248 	return sprintf(page, "%u\n", o2hb_dead_threshold);
2249 }
2250 
2251 static ssize_t o2hb_heartbeat_group_dead_threshold_store(struct config_item *item,
2252 		const char *page, size_t count)
2253 {
2254 	unsigned long tmp;
2255 	char *p = (char *)page;
2256 
2257 	tmp = simple_strtoul(p, &p, 10);
2258 	if (!p || (*p && (*p != '\n')))
2259                 return -EINVAL;
2260 
2261 	/* this will validate ranges for us. */
2262 	o2hb_dead_threshold_set((unsigned int) tmp);
2263 
2264 	return count;
2265 }
2266 
2267 static ssize_t o2hb_heartbeat_group_mode_show(struct config_item *item,
2268 		char *page)
2269 {
2270 	return sprintf(page, "%s\n",
2271 		       o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]);
2272 }
2273 
2274 static ssize_t o2hb_heartbeat_group_mode_store(struct config_item *item,
2275 		const char *page, size_t count)
2276 {
2277 	unsigned int i;
2278 	int ret;
2279 	size_t len;
2280 
2281 	len = (page[count - 1] == '\n') ? count - 1 : count;
2282 	if (!len)
2283 		return -EINVAL;
2284 
2285 	for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) {
2286 		if (strncasecmp(page, o2hb_heartbeat_mode_desc[i], len))
2287 			continue;
2288 
2289 		ret = o2hb_global_heartbeat_mode_set(i);
2290 		if (!ret)
2291 			printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n",
2292 			       o2hb_heartbeat_mode_desc[i]);
2293 		return count;
2294 	}
2295 
2296 	return -EINVAL;
2297 
2298 }
2299 
2300 CONFIGFS_ATTR(o2hb_heartbeat_group_, dead_threshold);
2301 CONFIGFS_ATTR(o2hb_heartbeat_group_, mode);
2302 
2303 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
2304 	&o2hb_heartbeat_group_attr_dead_threshold,
2305 	&o2hb_heartbeat_group_attr_mode,
2306 	NULL,
2307 };
2308 
2309 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
2310 	.make_item	= o2hb_heartbeat_group_make_item,
2311 	.drop_item	= o2hb_heartbeat_group_drop_item,
2312 };
2313 
2314 static struct config_item_type o2hb_heartbeat_group_type = {
2315 	.ct_group_ops	= &o2hb_heartbeat_group_group_ops,
2316 	.ct_attrs	= o2hb_heartbeat_group_attrs,
2317 	.ct_owner	= THIS_MODULE,
2318 };
2319 
2320 /* this is just here to avoid touching group in heartbeat.h which the
2321  * entire damn world #includes */
2322 struct config_group *o2hb_alloc_hb_set(void)
2323 {
2324 	struct o2hb_heartbeat_group *hs = NULL;
2325 	struct config_group *ret = NULL;
2326 
2327 	hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
2328 	if (hs == NULL)
2329 		goto out;
2330 
2331 	config_group_init_type_name(&hs->hs_group, "heartbeat",
2332 				    &o2hb_heartbeat_group_type);
2333 
2334 	ret = &hs->hs_group;
2335 out:
2336 	if (ret == NULL)
2337 		kfree(hs);
2338 	return ret;
2339 }
2340 
2341 void o2hb_free_hb_set(struct config_group *group)
2342 {
2343 	struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
2344 	kfree(hs);
2345 }
2346 
2347 /* hb callback registration and issuing */
2348 
2349 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
2350 {
2351 	if (type == O2HB_NUM_CB)
2352 		return ERR_PTR(-EINVAL);
2353 
2354 	return &o2hb_callbacks[type];
2355 }
2356 
2357 void o2hb_setup_callback(struct o2hb_callback_func *hc,
2358 			 enum o2hb_callback_type type,
2359 			 o2hb_cb_func *func,
2360 			 void *data,
2361 			 int priority)
2362 {
2363 	INIT_LIST_HEAD(&hc->hc_item);
2364 	hc->hc_func = func;
2365 	hc->hc_data = data;
2366 	hc->hc_priority = priority;
2367 	hc->hc_type = type;
2368 	hc->hc_magic = O2HB_CB_MAGIC;
2369 }
2370 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
2371 
2372 /*
2373  * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2374  * In global heartbeat mode, region_uuid passed is NULL.
2375  *
2376  * In local, we only pin the matching region. In global we pin all the active
2377  * regions.
2378  */
2379 static int o2hb_region_pin(const char *region_uuid)
2380 {
2381 	int ret = 0, found = 0;
2382 	struct o2hb_region *reg;
2383 	char *uuid;
2384 
2385 	assert_spin_locked(&o2hb_live_lock);
2386 
2387 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2388 		if (reg->hr_item_dropped)
2389 			continue;
2390 
2391 		uuid = config_item_name(&reg->hr_item);
2392 
2393 		/* local heartbeat */
2394 		if (region_uuid) {
2395 			if (strcmp(region_uuid, uuid))
2396 				continue;
2397 			found = 1;
2398 		}
2399 
2400 		if (reg->hr_item_pinned || reg->hr_item_dropped)
2401 			goto skip_pin;
2402 
2403 		/* Ignore ENOENT only for local hb (userdlm domain) */
2404 		ret = o2nm_depend_item(&reg->hr_item);
2405 		if (!ret) {
2406 			mlog(ML_CLUSTER, "Pin region %s\n", uuid);
2407 			reg->hr_item_pinned = 1;
2408 		} else {
2409 			if (ret == -ENOENT && found)
2410 				ret = 0;
2411 			else {
2412 				mlog(ML_ERROR, "Pin region %s fails with %d\n",
2413 				     uuid, ret);
2414 				break;
2415 			}
2416 		}
2417 skip_pin:
2418 		if (found)
2419 			break;
2420 	}
2421 
2422 	return ret;
2423 }
2424 
2425 /*
2426  * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2427  * In global heartbeat mode, region_uuid passed is NULL.
2428  *
2429  * In local, we only unpin the matching region. In global we unpin all the
2430  * active regions.
2431  */
2432 static void o2hb_region_unpin(const char *region_uuid)
2433 {
2434 	struct o2hb_region *reg;
2435 	char *uuid;
2436 	int found = 0;
2437 
2438 	assert_spin_locked(&o2hb_live_lock);
2439 
2440 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2441 		if (reg->hr_item_dropped)
2442 			continue;
2443 
2444 		uuid = config_item_name(&reg->hr_item);
2445 		if (region_uuid) {
2446 			if (strcmp(region_uuid, uuid))
2447 				continue;
2448 			found = 1;
2449 		}
2450 
2451 		if (reg->hr_item_pinned) {
2452 			mlog(ML_CLUSTER, "Unpin region %s\n", uuid);
2453 			o2nm_undepend_item(&reg->hr_item);
2454 			reg->hr_item_pinned = 0;
2455 		}
2456 		if (found)
2457 			break;
2458 	}
2459 }
2460 
2461 static int o2hb_region_inc_user(const char *region_uuid)
2462 {
2463 	int ret = 0;
2464 
2465 	spin_lock(&o2hb_live_lock);
2466 
2467 	/* local heartbeat */
2468 	if (!o2hb_global_heartbeat_active()) {
2469 	    ret = o2hb_region_pin(region_uuid);
2470 	    goto unlock;
2471 	}
2472 
2473 	/*
2474 	 * if global heartbeat active and this is the first dependent user,
2475 	 * pin all regions if quorum region count <= CUT_OFF
2476 	 */
2477 	o2hb_dependent_users++;
2478 	if (o2hb_dependent_users > 1)
2479 		goto unlock;
2480 
2481 	if (bitmap_weight(o2hb_quorum_region_bitmap,
2482 			   O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2483 		ret = o2hb_region_pin(NULL);
2484 
2485 unlock:
2486 	spin_unlock(&o2hb_live_lock);
2487 	return ret;
2488 }
2489 
2490 void o2hb_region_dec_user(const char *region_uuid)
2491 {
2492 	spin_lock(&o2hb_live_lock);
2493 
2494 	/* local heartbeat */
2495 	if (!o2hb_global_heartbeat_active()) {
2496 	    o2hb_region_unpin(region_uuid);
2497 	    goto unlock;
2498 	}
2499 
2500 	/*
2501 	 * if global heartbeat active and there are no dependent users,
2502 	 * unpin all quorum regions
2503 	 */
2504 	o2hb_dependent_users--;
2505 	if (!o2hb_dependent_users)
2506 		o2hb_region_unpin(NULL);
2507 
2508 unlock:
2509 	spin_unlock(&o2hb_live_lock);
2510 }
2511 
2512 int o2hb_register_callback(const char *region_uuid,
2513 			   struct o2hb_callback_func *hc)
2514 {
2515 	struct o2hb_callback_func *f;
2516 	struct o2hb_callback *hbcall;
2517 	int ret;
2518 
2519 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2520 	BUG_ON(!list_empty(&hc->hc_item));
2521 
2522 	hbcall = hbcall_from_type(hc->hc_type);
2523 	if (IS_ERR(hbcall)) {
2524 		ret = PTR_ERR(hbcall);
2525 		goto out;
2526 	}
2527 
2528 	if (region_uuid) {
2529 		ret = o2hb_region_inc_user(region_uuid);
2530 		if (ret) {
2531 			mlog_errno(ret);
2532 			goto out;
2533 		}
2534 	}
2535 
2536 	down_write(&o2hb_callback_sem);
2537 
2538 	list_for_each_entry(f, &hbcall->list, hc_item) {
2539 		if (hc->hc_priority < f->hc_priority) {
2540 			list_add_tail(&hc->hc_item, &f->hc_item);
2541 			break;
2542 		}
2543 	}
2544 	if (list_empty(&hc->hc_item))
2545 		list_add_tail(&hc->hc_item, &hbcall->list);
2546 
2547 	up_write(&o2hb_callback_sem);
2548 	ret = 0;
2549 out:
2550 	mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n",
2551 	     ret, __builtin_return_address(0), hc);
2552 	return ret;
2553 }
2554 EXPORT_SYMBOL_GPL(o2hb_register_callback);
2555 
2556 void o2hb_unregister_callback(const char *region_uuid,
2557 			      struct o2hb_callback_func *hc)
2558 {
2559 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2560 
2561 	mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n",
2562 	     __builtin_return_address(0), hc);
2563 
2564 	/* XXX Can this happen _with_ a region reference? */
2565 	if (list_empty(&hc->hc_item))
2566 		return;
2567 
2568 	if (region_uuid)
2569 		o2hb_region_dec_user(region_uuid);
2570 
2571 	down_write(&o2hb_callback_sem);
2572 
2573 	list_del_init(&hc->hc_item);
2574 
2575 	up_write(&o2hb_callback_sem);
2576 }
2577 EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
2578 
2579 int o2hb_check_node_heartbeating(u8 node_num)
2580 {
2581 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2582 
2583 	o2hb_fill_node_map(testing_map, sizeof(testing_map));
2584 	if (!test_bit(node_num, testing_map)) {
2585 		mlog(ML_HEARTBEAT,
2586 		     "node (%u) does not have heartbeating enabled.\n",
2587 		     node_num);
2588 		return 0;
2589 	}
2590 
2591 	return 1;
2592 }
2593 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating);
2594 
2595 int o2hb_check_node_heartbeating_no_sem(u8 node_num)
2596 {
2597 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2598 
2599 	spin_lock(&o2hb_live_lock);
2600 	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2601 	spin_unlock(&o2hb_live_lock);
2602 	if (!test_bit(node_num, testing_map)) {
2603 		mlog(ML_HEARTBEAT,
2604 		     "node (%u) does not have heartbeating enabled.\n",
2605 		     node_num);
2606 		return 0;
2607 	}
2608 
2609 	return 1;
2610 }
2611 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_no_sem);
2612 
2613 int o2hb_check_node_heartbeating_from_callback(u8 node_num)
2614 {
2615 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2616 
2617 	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2618 	if (!test_bit(node_num, testing_map)) {
2619 		mlog(ML_HEARTBEAT,
2620 		     "node (%u) does not have heartbeating enabled.\n",
2621 		     node_num);
2622 		return 0;
2623 	}
2624 
2625 	return 1;
2626 }
2627 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
2628 
2629 /* Makes sure our local node is configured with a node number, and is
2630  * heartbeating. */
2631 int o2hb_check_local_node_heartbeating(void)
2632 {
2633 	u8 node_num;
2634 
2635 	/* if this node was set then we have networking */
2636 	node_num = o2nm_this_node();
2637 	if (node_num == O2NM_MAX_NODES) {
2638 		mlog(ML_HEARTBEAT, "this node has not been configured.\n");
2639 		return 0;
2640 	}
2641 
2642 	return o2hb_check_node_heartbeating(node_num);
2643 }
2644 EXPORT_SYMBOL_GPL(o2hb_check_local_node_heartbeating);
2645 
2646 /*
2647  * this is just a hack until we get the plumbing which flips file systems
2648  * read only and drops the hb ref instead of killing the node dead.
2649  */
2650 void o2hb_stop_all_regions(void)
2651 {
2652 	struct o2hb_region *reg;
2653 
2654 	mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
2655 
2656 	spin_lock(&o2hb_live_lock);
2657 
2658 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
2659 		reg->hr_unclean_stop = 1;
2660 
2661 	spin_unlock(&o2hb_live_lock);
2662 }
2663 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
2664 
2665 int o2hb_get_all_regions(char *region_uuids, u8 max_regions)
2666 {
2667 	struct o2hb_region *reg;
2668 	int numregs = 0;
2669 	char *p;
2670 
2671 	spin_lock(&o2hb_live_lock);
2672 
2673 	p = region_uuids;
2674 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2675 		if (reg->hr_item_dropped)
2676 			continue;
2677 
2678 		mlog(0, "Region: %s\n", config_item_name(&reg->hr_item));
2679 		if (numregs < max_regions) {
2680 			memcpy(p, config_item_name(&reg->hr_item),
2681 			       O2HB_MAX_REGION_NAME_LEN);
2682 			p += O2HB_MAX_REGION_NAME_LEN;
2683 		}
2684 		numregs++;
2685 	}
2686 
2687 	spin_unlock(&o2hb_live_lock);
2688 
2689 	return numregs;
2690 }
2691 EXPORT_SYMBOL_GPL(o2hb_get_all_regions);
2692 
2693 int o2hb_global_heartbeat_active(void)
2694 {
2695 	return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL);
2696 }
2697 EXPORT_SYMBOL(o2hb_global_heartbeat_active);
2698