heartbeat.c

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