xref: /openbmc/linux/fs/ocfs2/cluster/heartbeat.c (revision 64c70b1c)
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 
37 #include "heartbeat.h"
38 #include "tcp.h"
39 #include "nodemanager.h"
40 #include "quorum.h"
41 
42 #include "masklog.h"
43 
44 
45 /*
46  * The first heartbeat pass had one global thread that would serialize all hb
47  * callback calls.  This global serializing sem should only be removed once
48  * we've made sure that all callees can deal with being called concurrently
49  * from multiple hb region threads.
50  */
51 static DECLARE_RWSEM(o2hb_callback_sem);
52 
53 /*
54  * multiple hb threads are watching multiple regions.  A node is live
55  * whenever any of the threads sees activity from the node in its region.
56  */
57 static DEFINE_SPINLOCK(o2hb_live_lock);
58 static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
59 static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
60 static LIST_HEAD(o2hb_node_events);
61 static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
62 
63 static LIST_HEAD(o2hb_all_regions);
64 
65 static struct o2hb_callback {
66 	struct list_head list;
67 } o2hb_callbacks[O2HB_NUM_CB];
68 
69 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
70 
71 #define O2HB_DEFAULT_BLOCK_BITS       9
72 
73 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
74 
75 /* Only sets a new threshold if there are no active regions.
76  *
77  * No locking or otherwise interesting code is required for reading
78  * o2hb_dead_threshold as it can't change once regions are active and
79  * it's not interesting to anyone until then anyway. */
80 static void o2hb_dead_threshold_set(unsigned int threshold)
81 {
82 	if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
83 		spin_lock(&o2hb_live_lock);
84 		if (list_empty(&o2hb_all_regions))
85 			o2hb_dead_threshold = threshold;
86 		spin_unlock(&o2hb_live_lock);
87 	}
88 }
89 
90 struct o2hb_node_event {
91 	struct list_head        hn_item;
92 	enum o2hb_callback_type hn_event_type;
93 	struct o2nm_node        *hn_node;
94 	int                     hn_node_num;
95 };
96 
97 struct o2hb_disk_slot {
98 	struct o2hb_disk_heartbeat_block *ds_raw_block;
99 	u8			ds_node_num;
100 	u64			ds_last_time;
101 	u64			ds_last_generation;
102 	u16			ds_equal_samples;
103 	u16			ds_changed_samples;
104 	struct list_head	ds_live_item;
105 };
106 
107 /* each thread owns a region.. when we're asked to tear down the region
108  * we ask the thread to stop, who cleans up the region */
109 struct o2hb_region {
110 	struct config_item	hr_item;
111 
112 	struct list_head	hr_all_item;
113 	unsigned		hr_unclean_stop:1;
114 
115 	/* protected by the hr_callback_sem */
116 	struct task_struct 	*hr_task;
117 
118 	unsigned int		hr_blocks;
119 	unsigned long long	hr_start_block;
120 
121 	unsigned int		hr_block_bits;
122 	unsigned int		hr_block_bytes;
123 
124 	unsigned int		hr_slots_per_page;
125 	unsigned int		hr_num_pages;
126 
127 	struct page             **hr_slot_data;
128 	struct block_device	*hr_bdev;
129 	struct o2hb_disk_slot	*hr_slots;
130 
131 	/* let the person setting up hb wait for it to return until it
132 	 * has reached a 'steady' state.  This will be fixed when we have
133 	 * a more complete api that doesn't lead to this sort of fragility. */
134 	atomic_t		hr_steady_iterations;
135 
136 	char			hr_dev_name[BDEVNAME_SIZE];
137 
138 	unsigned int		hr_timeout_ms;
139 
140 	/* randomized as the region goes up and down so that a node
141 	 * recognizes a node going up and down in one iteration */
142 	u64			hr_generation;
143 
144 	struct delayed_work	hr_write_timeout_work;
145 	unsigned long		hr_last_timeout_start;
146 
147 	/* Used during o2hb_check_slot to hold a copy of the block
148 	 * being checked because we temporarily have to zero out the
149 	 * crc field. */
150 	struct o2hb_disk_heartbeat_block *hr_tmp_block;
151 };
152 
153 struct o2hb_bio_wait_ctxt {
154 	atomic_t          wc_num_reqs;
155 	struct completion wc_io_complete;
156 	int               wc_error;
157 };
158 
159 static void o2hb_write_timeout(struct work_struct *work)
160 {
161 	struct o2hb_region *reg =
162 		container_of(work, struct o2hb_region,
163 			     hr_write_timeout_work.work);
164 
165 	mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
166 	     "milliseconds\n", reg->hr_dev_name,
167 	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
168 	o2quo_disk_timeout();
169 }
170 
171 static void o2hb_arm_write_timeout(struct o2hb_region *reg)
172 {
173 	mlog(0, "Queue write timeout for %u ms\n", O2HB_MAX_WRITE_TIMEOUT_MS);
174 
175 	cancel_delayed_work(&reg->hr_write_timeout_work);
176 	reg->hr_last_timeout_start = jiffies;
177 	schedule_delayed_work(&reg->hr_write_timeout_work,
178 			      msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
179 }
180 
181 static void o2hb_disarm_write_timeout(struct o2hb_region *reg)
182 {
183 	cancel_delayed_work(&reg->hr_write_timeout_work);
184 	flush_scheduled_work();
185 }
186 
187 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc)
188 {
189 	atomic_set(&wc->wc_num_reqs, 1);
190 	init_completion(&wc->wc_io_complete);
191 	wc->wc_error = 0;
192 }
193 
194 /* Used in error paths too */
195 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
196 				     unsigned int num)
197 {
198 	/* sadly atomic_sub_and_test() isn't available on all platforms.  The
199 	 * good news is that the fast path only completes one at a time */
200 	while(num--) {
201 		if (atomic_dec_and_test(&wc->wc_num_reqs)) {
202 			BUG_ON(num > 0);
203 			complete(&wc->wc_io_complete);
204 		}
205 	}
206 }
207 
208 static void o2hb_wait_on_io(struct o2hb_region *reg,
209 			    struct o2hb_bio_wait_ctxt *wc)
210 {
211 	struct address_space *mapping = reg->hr_bdev->bd_inode->i_mapping;
212 
213 	blk_run_address_space(mapping);
214 	o2hb_bio_wait_dec(wc, 1);
215 
216 	wait_for_completion(&wc->wc_io_complete);
217 }
218 
219 static int o2hb_bio_end_io(struct bio *bio,
220 			   unsigned int bytes_done,
221 			   int error)
222 {
223 	struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
224 
225 	if (error) {
226 		mlog(ML_ERROR, "IO Error %d\n", error);
227 		wc->wc_error = error;
228 	}
229 
230 	if (bio->bi_size)
231 		return 1;
232 
233 	o2hb_bio_wait_dec(wc, 1);
234 	bio_put(bio);
235 	return 0;
236 }
237 
238 /* Setup a Bio to cover I/O against num_slots slots starting at
239  * start_slot. */
240 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
241 				      struct o2hb_bio_wait_ctxt *wc,
242 				      unsigned int *current_slot,
243 				      unsigned int max_slots)
244 {
245 	int len, current_page;
246 	unsigned int vec_len, vec_start;
247 	unsigned int bits = reg->hr_block_bits;
248 	unsigned int spp = reg->hr_slots_per_page;
249 	unsigned int cs = *current_slot;
250 	struct bio *bio;
251 	struct page *page;
252 
253 	/* Testing has shown this allocation to take long enough under
254 	 * GFP_KERNEL that the local node can get fenced. It would be
255 	 * nicest if we could pre-allocate these bios and avoid this
256 	 * all together. */
257 	bio = bio_alloc(GFP_ATOMIC, 16);
258 	if (!bio) {
259 		mlog(ML_ERROR, "Could not alloc slots BIO!\n");
260 		bio = ERR_PTR(-ENOMEM);
261 		goto bail;
262 	}
263 
264 	/* Must put everything in 512 byte sectors for the bio... */
265 	bio->bi_sector = (reg->hr_start_block + cs) << (bits - 9);
266 	bio->bi_bdev = reg->hr_bdev;
267 	bio->bi_private = wc;
268 	bio->bi_end_io = o2hb_bio_end_io;
269 
270 	vec_start = (cs << bits) % PAGE_CACHE_SIZE;
271 	while(cs < max_slots) {
272 		current_page = cs / spp;
273 		page = reg->hr_slot_data[current_page];
274 
275 		vec_len = min(PAGE_CACHE_SIZE,
276 			      (max_slots-cs) * (PAGE_CACHE_SIZE/spp) );
277 
278 		mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
279 		     current_page, vec_len, vec_start);
280 
281 		len = bio_add_page(bio, page, vec_len, vec_start);
282 		if (len != vec_len) break;
283 
284 		cs += vec_len / (PAGE_CACHE_SIZE/spp);
285 		vec_start = 0;
286 	}
287 
288 bail:
289 	*current_slot = cs;
290 	return bio;
291 }
292 
293 static int o2hb_read_slots(struct o2hb_region *reg,
294 			   unsigned int max_slots)
295 {
296 	unsigned int current_slot=0;
297 	int status;
298 	struct o2hb_bio_wait_ctxt wc;
299 	struct bio *bio;
300 
301 	o2hb_bio_wait_init(&wc);
302 
303 	while(current_slot < max_slots) {
304 		bio = o2hb_setup_one_bio(reg, &wc, &current_slot, max_slots);
305 		if (IS_ERR(bio)) {
306 			status = PTR_ERR(bio);
307 			mlog_errno(status);
308 			goto bail_and_wait;
309 		}
310 
311 		atomic_inc(&wc.wc_num_reqs);
312 		submit_bio(READ, bio);
313 	}
314 
315 	status = 0;
316 
317 bail_and_wait:
318 	o2hb_wait_on_io(reg, &wc);
319 	if (wc.wc_error && !status)
320 		status = wc.wc_error;
321 
322 	return status;
323 }
324 
325 static int o2hb_issue_node_write(struct o2hb_region *reg,
326 				 struct o2hb_bio_wait_ctxt *write_wc)
327 {
328 	int status;
329 	unsigned int slot;
330 	struct bio *bio;
331 
332 	o2hb_bio_wait_init(write_wc);
333 
334 	slot = o2nm_this_node();
335 
336 	bio = o2hb_setup_one_bio(reg, write_wc, &slot, slot+1);
337 	if (IS_ERR(bio)) {
338 		status = PTR_ERR(bio);
339 		mlog_errno(status);
340 		goto bail;
341 	}
342 
343 	atomic_inc(&write_wc->wc_num_reqs);
344 	submit_bio(WRITE, bio);
345 
346 	status = 0;
347 bail:
348 	return status;
349 }
350 
351 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
352 				     struct o2hb_disk_heartbeat_block *hb_block)
353 {
354 	__le32 old_cksum;
355 	u32 ret;
356 
357 	/* We want to compute the block crc with a 0 value in the
358 	 * hb_cksum field. Save it off here and replace after the
359 	 * crc. */
360 	old_cksum = hb_block->hb_cksum;
361 	hb_block->hb_cksum = 0;
362 
363 	ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
364 
365 	hb_block->hb_cksum = old_cksum;
366 
367 	return ret;
368 }
369 
370 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
371 {
372 	mlog(ML_ERROR, "Dump slot information: seq = 0x%llx, node = %u, "
373 	     "cksum = 0x%x, generation 0x%llx\n",
374 	     (long long)le64_to_cpu(hb_block->hb_seq),
375 	     hb_block->hb_node, le32_to_cpu(hb_block->hb_cksum),
376 	     (long long)le64_to_cpu(hb_block->hb_generation));
377 }
378 
379 static int o2hb_verify_crc(struct o2hb_region *reg,
380 			   struct o2hb_disk_heartbeat_block *hb_block)
381 {
382 	u32 read, computed;
383 
384 	read = le32_to_cpu(hb_block->hb_cksum);
385 	computed = o2hb_compute_block_crc_le(reg, hb_block);
386 
387 	return read == computed;
388 }
389 
390 /* We want to make sure that nobody is heartbeating on top of us --
391  * this will help detect an invalid configuration. */
392 static int o2hb_check_last_timestamp(struct o2hb_region *reg)
393 {
394 	int node_num, ret;
395 	struct o2hb_disk_slot *slot;
396 	struct o2hb_disk_heartbeat_block *hb_block;
397 
398 	node_num = o2nm_this_node();
399 
400 	ret = 1;
401 	slot = &reg->hr_slots[node_num];
402 	/* Don't check on our 1st timestamp */
403 	if (slot->ds_last_time) {
404 		hb_block = slot->ds_raw_block;
405 
406 		if (le64_to_cpu(hb_block->hb_seq) != slot->ds_last_time)
407 			ret = 0;
408 	}
409 
410 	return ret;
411 }
412 
413 static inline void o2hb_prepare_block(struct o2hb_region *reg,
414 				      u64 generation)
415 {
416 	int node_num;
417 	u64 cputime;
418 	struct o2hb_disk_slot *slot;
419 	struct o2hb_disk_heartbeat_block *hb_block;
420 
421 	node_num = o2nm_this_node();
422 	slot = &reg->hr_slots[node_num];
423 
424 	hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
425 	memset(hb_block, 0, reg->hr_block_bytes);
426 	/* TODO: time stuff */
427 	cputime = CURRENT_TIME.tv_sec;
428 	if (!cputime)
429 		cputime = 1;
430 
431 	hb_block->hb_seq = cpu_to_le64(cputime);
432 	hb_block->hb_node = node_num;
433 	hb_block->hb_generation = cpu_to_le64(generation);
434 	hb_block->hb_dead_ms = cpu_to_le32(o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS);
435 
436 	/* This step must always happen last! */
437 	hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
438 								   hb_block));
439 
440 	mlog(ML_HB_BIO, "our node generation = 0x%llx, cksum = 0x%x\n",
441 	     (long long)generation,
442 	     le32_to_cpu(hb_block->hb_cksum));
443 }
444 
445 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
446 				struct o2nm_node *node,
447 				int idx)
448 {
449 	struct list_head *iter;
450 	struct o2hb_callback_func *f;
451 
452 	list_for_each(iter, &hbcall->list) {
453 		f = list_entry(iter, struct o2hb_callback_func, hc_item);
454 		mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
455 		(f->hc_func)(node, idx, f->hc_data);
456 	}
457 }
458 
459 /* Will run the list in order until we process the passed event */
460 static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
461 {
462 	int empty;
463 	struct o2hb_callback *hbcall;
464 	struct o2hb_node_event *event;
465 
466 	spin_lock(&o2hb_live_lock);
467 	empty = list_empty(&queued_event->hn_item);
468 	spin_unlock(&o2hb_live_lock);
469 	if (empty)
470 		return;
471 
472 	/* Holding callback sem assures we don't alter the callback
473 	 * lists when doing this, and serializes ourselves with other
474 	 * processes wanting callbacks. */
475 	down_write(&o2hb_callback_sem);
476 
477 	spin_lock(&o2hb_live_lock);
478 	while (!list_empty(&o2hb_node_events)
479 	       && !list_empty(&queued_event->hn_item)) {
480 		event = list_entry(o2hb_node_events.next,
481 				   struct o2hb_node_event,
482 				   hn_item);
483 		list_del_init(&event->hn_item);
484 		spin_unlock(&o2hb_live_lock);
485 
486 		mlog(ML_HEARTBEAT, "Node %s event for %d\n",
487 		     event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
488 		     event->hn_node_num);
489 
490 		hbcall = hbcall_from_type(event->hn_event_type);
491 
492 		/* We should *never* have gotten on to the list with a
493 		 * bad type... This isn't something that we should try
494 		 * to recover from. */
495 		BUG_ON(IS_ERR(hbcall));
496 
497 		o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
498 
499 		spin_lock(&o2hb_live_lock);
500 	}
501 	spin_unlock(&o2hb_live_lock);
502 
503 	up_write(&o2hb_callback_sem);
504 }
505 
506 static void o2hb_queue_node_event(struct o2hb_node_event *event,
507 				  enum o2hb_callback_type type,
508 				  struct o2nm_node *node,
509 				  int node_num)
510 {
511 	assert_spin_locked(&o2hb_live_lock);
512 
513 	event->hn_event_type = type;
514 	event->hn_node = node;
515 	event->hn_node_num = node_num;
516 
517 	mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
518 	     type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
519 
520 	list_add_tail(&event->hn_item, &o2hb_node_events);
521 }
522 
523 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
524 {
525 	struct o2hb_node_event event =
526 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
527 	struct o2nm_node *node;
528 
529 	node = o2nm_get_node_by_num(slot->ds_node_num);
530 	if (!node)
531 		return;
532 
533 	spin_lock(&o2hb_live_lock);
534 	if (!list_empty(&slot->ds_live_item)) {
535 		mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
536 		     slot->ds_node_num);
537 
538 		list_del_init(&slot->ds_live_item);
539 
540 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
541 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
542 
543 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
544 					      slot->ds_node_num);
545 		}
546 	}
547 	spin_unlock(&o2hb_live_lock);
548 
549 	o2hb_run_event_list(&event);
550 
551 	o2nm_node_put(node);
552 }
553 
554 static int o2hb_check_slot(struct o2hb_region *reg,
555 			   struct o2hb_disk_slot *slot)
556 {
557 	int changed = 0, gen_changed = 0;
558 	struct o2hb_node_event event =
559 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
560 	struct o2nm_node *node;
561 	struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
562 	u64 cputime;
563 	unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS;
564 	unsigned int slot_dead_ms;
565 
566 	memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
567 
568 	/* Is this correct? Do we assume that the node doesn't exist
569 	 * if we're not configured for him? */
570 	node = o2nm_get_node_by_num(slot->ds_node_num);
571 	if (!node)
572 		return 0;
573 
574 	if (!o2hb_verify_crc(reg, hb_block)) {
575 		/* all paths from here will drop o2hb_live_lock for
576 		 * us. */
577 		spin_lock(&o2hb_live_lock);
578 
579 		/* Don't print an error on the console in this case -
580 		 * a freshly formatted heartbeat area will not have a
581 		 * crc set on it. */
582 		if (list_empty(&slot->ds_live_item))
583 			goto out;
584 
585 		/* The node is live but pushed out a bad crc. We
586 		 * consider it a transient miss but don't populate any
587 		 * other values as they may be junk. */
588 		mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
589 		     slot->ds_node_num, reg->hr_dev_name);
590 		o2hb_dump_slot(hb_block);
591 
592 		slot->ds_equal_samples++;
593 		goto fire_callbacks;
594 	}
595 
596 	/* we don't care if these wrap.. the state transitions below
597 	 * clear at the right places */
598 	cputime = le64_to_cpu(hb_block->hb_seq);
599 	if (slot->ds_last_time != cputime)
600 		slot->ds_changed_samples++;
601 	else
602 		slot->ds_equal_samples++;
603 	slot->ds_last_time = cputime;
604 
605 	/* The node changed heartbeat generations. We assume this to
606 	 * mean it dropped off but came back before we timed out. We
607 	 * want to consider it down for the time being but don't want
608 	 * to lose any changed_samples state we might build up to
609 	 * considering it live again. */
610 	if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
611 		gen_changed = 1;
612 		slot->ds_equal_samples = 0;
613 		mlog(ML_HEARTBEAT, "Node %d changed generation (0x%llx "
614 		     "to 0x%llx)\n", slot->ds_node_num,
615 		     (long long)slot->ds_last_generation,
616 		     (long long)le64_to_cpu(hb_block->hb_generation));
617 	}
618 
619 	slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
620 
621 	mlog(ML_HEARTBEAT, "Slot %d gen 0x%llx cksum 0x%x "
622 	     "seq %llu last %llu changed %u equal %u\n",
623 	     slot->ds_node_num, (long long)slot->ds_last_generation,
624 	     le32_to_cpu(hb_block->hb_cksum),
625 	     (unsigned long long)le64_to_cpu(hb_block->hb_seq),
626 	     (unsigned long long)slot->ds_last_time, slot->ds_changed_samples,
627 	     slot->ds_equal_samples);
628 
629 	spin_lock(&o2hb_live_lock);
630 
631 fire_callbacks:
632 	/* dead nodes only come to life after some number of
633 	 * changes at any time during their dead time */
634 	if (list_empty(&slot->ds_live_item) &&
635 	    slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
636 		mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n",
637 		     slot->ds_node_num, (long long)slot->ds_last_generation);
638 
639 		/* first on the list generates a callback */
640 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
641 			set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
642 
643 			o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
644 					      slot->ds_node_num);
645 
646 			changed = 1;
647 		}
648 
649 		list_add_tail(&slot->ds_live_item,
650 			      &o2hb_live_slots[slot->ds_node_num]);
651 
652 		slot->ds_equal_samples = 0;
653 
654 		/* We want to be sure that all nodes agree on the
655 		 * number of milliseconds before a node will be
656 		 * considered dead. The self-fencing timeout is
657 		 * computed from this value, and a discrepancy might
658 		 * result in heartbeat calling a node dead when it
659 		 * hasn't self-fenced yet. */
660 		slot_dead_ms = le32_to_cpu(hb_block->hb_dead_ms);
661 		if (slot_dead_ms && slot_dead_ms != dead_ms) {
662 			/* TODO: Perhaps we can fail the region here. */
663 			mlog(ML_ERROR, "Node %d on device %s has a dead count "
664 			     "of %u ms, but our count is %u ms.\n"
665 			     "Please double check your configuration values "
666 			     "for 'O2CB_HEARTBEAT_THRESHOLD'\n",
667 			     slot->ds_node_num, reg->hr_dev_name, slot_dead_ms,
668 			     dead_ms);
669 		}
670 		goto out;
671 	}
672 
673 	/* if the list is dead, we're done.. */
674 	if (list_empty(&slot->ds_live_item))
675 		goto out;
676 
677 	/* live nodes only go dead after enough consequtive missed
678 	 * samples..  reset the missed counter whenever we see
679 	 * activity */
680 	if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
681 		mlog(ML_HEARTBEAT, "Node %d left my region\n",
682 		     slot->ds_node_num);
683 
684 		/* last off the live_slot generates a callback */
685 		list_del_init(&slot->ds_live_item);
686 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
687 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
688 
689 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
690 					      slot->ds_node_num);
691 
692 			changed = 1;
693 		}
694 
695 		/* We don't clear this because the node is still
696 		 * actually writing new blocks. */
697 		if (!gen_changed)
698 			slot->ds_changed_samples = 0;
699 		goto out;
700 	}
701 	if (slot->ds_changed_samples) {
702 		slot->ds_changed_samples = 0;
703 		slot->ds_equal_samples = 0;
704 	}
705 out:
706 	spin_unlock(&o2hb_live_lock);
707 
708 	o2hb_run_event_list(&event);
709 
710 	o2nm_node_put(node);
711 	return changed;
712 }
713 
714 /* This could be faster if we just implmented a find_last_bit, but I
715  * don't think the circumstances warrant it. */
716 static int o2hb_highest_node(unsigned long *nodes,
717 			     int numbits)
718 {
719 	int highest, node;
720 
721 	highest = numbits;
722 	node = -1;
723 	while ((node = find_next_bit(nodes, numbits, node + 1)) != -1) {
724 		if (node >= numbits)
725 			break;
726 
727 		highest = node;
728 	}
729 
730 	return highest;
731 }
732 
733 static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
734 {
735 	int i, ret, highest_node, change = 0;
736 	unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
737 	struct o2hb_bio_wait_ctxt write_wc;
738 
739 	ret = o2nm_configured_node_map(configured_nodes,
740 				       sizeof(configured_nodes));
741 	if (ret) {
742 		mlog_errno(ret);
743 		return ret;
744 	}
745 
746 	highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
747 	if (highest_node >= O2NM_MAX_NODES) {
748 		mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n");
749 		return -EINVAL;
750 	}
751 
752 	/* No sense in reading the slots of nodes that don't exist
753 	 * yet. Of course, if the node definitions have holes in them
754 	 * then we're reading an empty slot anyway... Consider this
755 	 * best-effort. */
756 	ret = o2hb_read_slots(reg, highest_node + 1);
757 	if (ret < 0) {
758 		mlog_errno(ret);
759 		return ret;
760 	}
761 
762 	/* With an up to date view of the slots, we can check that no
763 	 * other node has been improperly configured to heartbeat in
764 	 * our slot. */
765 	if (!o2hb_check_last_timestamp(reg))
766 		mlog(ML_ERROR, "Device \"%s\": another node is heartbeating "
767 		     "in our slot!\n", reg->hr_dev_name);
768 
769 	/* fill in the proper info for our next heartbeat */
770 	o2hb_prepare_block(reg, reg->hr_generation);
771 
772 	/* And fire off the write. Note that we don't wait on this I/O
773 	 * until later. */
774 	ret = o2hb_issue_node_write(reg, &write_wc);
775 	if (ret < 0) {
776 		mlog_errno(ret);
777 		return ret;
778 	}
779 
780 	i = -1;
781 	while((i = find_next_bit(configured_nodes, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
782 
783 		change |= o2hb_check_slot(reg, &reg->hr_slots[i]);
784 	}
785 
786 	/*
787 	 * We have to be sure we've advertised ourselves on disk
788 	 * before we can go to steady state.  This ensures that
789 	 * people we find in our steady state have seen us.
790 	 */
791 	o2hb_wait_on_io(reg, &write_wc);
792 	if (write_wc.wc_error) {
793 		/* Do not re-arm the write timeout on I/O error - we
794 		 * can't be sure that the new block ever made it to
795 		 * disk */
796 		mlog(ML_ERROR, "Write error %d on device \"%s\"\n",
797 		     write_wc.wc_error, reg->hr_dev_name);
798 		return write_wc.wc_error;
799 	}
800 
801 	o2hb_arm_write_timeout(reg);
802 
803 	/* let the person who launched us know when things are steady */
804 	if (!change && (atomic_read(&reg->hr_steady_iterations) != 0)) {
805 		if (atomic_dec_and_test(&reg->hr_steady_iterations))
806 			wake_up(&o2hb_steady_queue);
807 	}
808 
809 	return 0;
810 }
811 
812 /* Subtract b from a, storing the result in a. a *must* have a larger
813  * value than b. */
814 static void o2hb_tv_subtract(struct timeval *a,
815 			     struct timeval *b)
816 {
817 	/* just return 0 when a is after b */
818 	if (a->tv_sec < b->tv_sec ||
819 	    (a->tv_sec == b->tv_sec && a->tv_usec < b->tv_usec)) {
820 		a->tv_sec = 0;
821 		a->tv_usec = 0;
822 		return;
823 	}
824 
825 	a->tv_sec -= b->tv_sec;
826 	a->tv_usec -= b->tv_usec;
827 	while ( a->tv_usec < 0 ) {
828 		a->tv_sec--;
829 		a->tv_usec += 1000000;
830 	}
831 }
832 
833 static unsigned int o2hb_elapsed_msecs(struct timeval *start,
834 				       struct timeval *end)
835 {
836 	struct timeval res = *end;
837 
838 	o2hb_tv_subtract(&res, start);
839 
840 	return res.tv_sec * 1000 + res.tv_usec / 1000;
841 }
842 
843 /*
844  * we ride the region ref that the region dir holds.  before the region
845  * dir is removed and drops it ref it will wait to tear down this
846  * thread.
847  */
848 static int o2hb_thread(void *data)
849 {
850 	int i, ret;
851 	struct o2hb_region *reg = data;
852 	struct o2hb_bio_wait_ctxt write_wc;
853 	struct timeval before_hb, after_hb;
854 	unsigned int elapsed_msec;
855 
856 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
857 
858 	set_user_nice(current, -20);
859 
860 	while (!kthread_should_stop() && !reg->hr_unclean_stop) {
861 		/* We track the time spent inside
862 		 * o2hb_do_disk_heartbeat so that we avoid more then
863 		 * hr_timeout_ms between disk writes. On busy systems
864 		 * this should result in a heartbeat which is less
865 		 * likely to time itself out. */
866 		do_gettimeofday(&before_hb);
867 
868 		i = 0;
869 		do {
870 			ret = o2hb_do_disk_heartbeat(reg);
871 		} while (ret && ++i < 2);
872 
873 		do_gettimeofday(&after_hb);
874 		elapsed_msec = o2hb_elapsed_msecs(&before_hb, &after_hb);
875 
876 		mlog(0, "start = %lu.%lu, end = %lu.%lu, msec = %u\n",
877 		     before_hb.tv_sec, (unsigned long) before_hb.tv_usec,
878 		     after_hb.tv_sec, (unsigned long) after_hb.tv_usec,
879 		     elapsed_msec);
880 
881 		if (elapsed_msec < reg->hr_timeout_ms) {
882 			/* the kthread api has blocked signals for us so no
883 			 * need to record the return value. */
884 			msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
885 		}
886 	}
887 
888 	o2hb_disarm_write_timeout(reg);
889 
890 	/* unclean stop is only used in very bad situation */
891 	for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
892 		o2hb_shutdown_slot(&reg->hr_slots[i]);
893 
894 	/* Explicit down notification - avoid forcing the other nodes
895 	 * to timeout on this region when we could just as easily
896 	 * write a clear generation - thus indicating to them that
897 	 * this node has left this region.
898 	 *
899 	 * XXX: Should we skip this on unclean_stop? */
900 	o2hb_prepare_block(reg, 0);
901 	ret = o2hb_issue_node_write(reg, &write_wc);
902 	if (ret == 0) {
903 		o2hb_wait_on_io(reg, &write_wc);
904 	} else {
905 		mlog_errno(ret);
906 	}
907 
908 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n");
909 
910 	return 0;
911 }
912 
913 void o2hb_init(void)
914 {
915 	int i;
916 
917 	for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
918 		INIT_LIST_HEAD(&o2hb_callbacks[i].list);
919 
920 	for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
921 		INIT_LIST_HEAD(&o2hb_live_slots[i]);
922 
923 	INIT_LIST_HEAD(&o2hb_node_events);
924 
925 	memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
926 }
927 
928 /* if we're already in a callback then we're already serialized by the sem */
929 static void o2hb_fill_node_map_from_callback(unsigned long *map,
930 					     unsigned bytes)
931 {
932 	BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
933 
934 	memcpy(map, &o2hb_live_node_bitmap, bytes);
935 }
936 
937 /*
938  * get a map of all nodes that are heartbeating in any regions
939  */
940 void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
941 {
942 	/* callers want to serialize this map and callbacks so that they
943 	 * can trust that they don't miss nodes coming to the party */
944 	down_read(&o2hb_callback_sem);
945 	spin_lock(&o2hb_live_lock);
946 	o2hb_fill_node_map_from_callback(map, bytes);
947 	spin_unlock(&o2hb_live_lock);
948 	up_read(&o2hb_callback_sem);
949 }
950 EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
951 
952 /*
953  * heartbeat configfs bits.  The heartbeat set is a default set under
954  * the cluster set in nodemanager.c.
955  */
956 
957 static struct o2hb_region *to_o2hb_region(struct config_item *item)
958 {
959 	return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
960 }
961 
962 /* drop_item only drops its ref after killing the thread, nothing should
963  * be using the region anymore.  this has to clean up any state that
964  * attributes might have built up. */
965 static void o2hb_region_release(struct config_item *item)
966 {
967 	int i;
968 	struct page *page;
969 	struct o2hb_region *reg = to_o2hb_region(item);
970 
971 	if (reg->hr_tmp_block)
972 		kfree(reg->hr_tmp_block);
973 
974 	if (reg->hr_slot_data) {
975 		for (i = 0; i < reg->hr_num_pages; i++) {
976 			page = reg->hr_slot_data[i];
977 			if (page)
978 				__free_page(page);
979 		}
980 		kfree(reg->hr_slot_data);
981 	}
982 
983 	if (reg->hr_bdev)
984 		blkdev_put(reg->hr_bdev);
985 
986 	if (reg->hr_slots)
987 		kfree(reg->hr_slots);
988 
989 	spin_lock(&o2hb_live_lock);
990 	list_del(&reg->hr_all_item);
991 	spin_unlock(&o2hb_live_lock);
992 
993 	kfree(reg);
994 }
995 
996 static int o2hb_read_block_input(struct o2hb_region *reg,
997 				 const char *page,
998 				 size_t count,
999 				 unsigned long *ret_bytes,
1000 				 unsigned int *ret_bits)
1001 {
1002 	unsigned long bytes;
1003 	char *p = (char *)page;
1004 
1005 	bytes = simple_strtoul(p, &p, 0);
1006 	if (!p || (*p && (*p != '\n')))
1007 		return -EINVAL;
1008 
1009 	/* Heartbeat and fs min / max block sizes are the same. */
1010 	if (bytes > 4096 || bytes < 512)
1011 		return -ERANGE;
1012 	if (hweight16(bytes) != 1)
1013 		return -EINVAL;
1014 
1015 	if (ret_bytes)
1016 		*ret_bytes = bytes;
1017 	if (ret_bits)
1018 		*ret_bits = ffs(bytes) - 1;
1019 
1020 	return 0;
1021 }
1022 
1023 static ssize_t o2hb_region_block_bytes_read(struct o2hb_region *reg,
1024 					    char *page)
1025 {
1026 	return sprintf(page, "%u\n", reg->hr_block_bytes);
1027 }
1028 
1029 static ssize_t o2hb_region_block_bytes_write(struct o2hb_region *reg,
1030 					     const char *page,
1031 					     size_t count)
1032 {
1033 	int status;
1034 	unsigned long block_bytes;
1035 	unsigned int block_bits;
1036 
1037 	if (reg->hr_bdev)
1038 		return -EINVAL;
1039 
1040 	status = o2hb_read_block_input(reg, page, count,
1041 				       &block_bytes, &block_bits);
1042 	if (status)
1043 		return status;
1044 
1045 	reg->hr_block_bytes = (unsigned int)block_bytes;
1046 	reg->hr_block_bits = block_bits;
1047 
1048 	return count;
1049 }
1050 
1051 static ssize_t o2hb_region_start_block_read(struct o2hb_region *reg,
1052 					    char *page)
1053 {
1054 	return sprintf(page, "%llu\n", reg->hr_start_block);
1055 }
1056 
1057 static ssize_t o2hb_region_start_block_write(struct o2hb_region *reg,
1058 					     const char *page,
1059 					     size_t count)
1060 {
1061 	unsigned long long tmp;
1062 	char *p = (char *)page;
1063 
1064 	if (reg->hr_bdev)
1065 		return -EINVAL;
1066 
1067 	tmp = simple_strtoull(p, &p, 0);
1068 	if (!p || (*p && (*p != '\n')))
1069 		return -EINVAL;
1070 
1071 	reg->hr_start_block = tmp;
1072 
1073 	return count;
1074 }
1075 
1076 static ssize_t o2hb_region_blocks_read(struct o2hb_region *reg,
1077 				       char *page)
1078 {
1079 	return sprintf(page, "%d\n", reg->hr_blocks);
1080 }
1081 
1082 static ssize_t o2hb_region_blocks_write(struct o2hb_region *reg,
1083 					const char *page,
1084 					size_t count)
1085 {
1086 	unsigned long tmp;
1087 	char *p = (char *)page;
1088 
1089 	if (reg->hr_bdev)
1090 		return -EINVAL;
1091 
1092 	tmp = simple_strtoul(p, &p, 0);
1093 	if (!p || (*p && (*p != '\n')))
1094 		return -EINVAL;
1095 
1096 	if (tmp > O2NM_MAX_NODES || tmp == 0)
1097 		return -ERANGE;
1098 
1099 	reg->hr_blocks = (unsigned int)tmp;
1100 
1101 	return count;
1102 }
1103 
1104 static ssize_t o2hb_region_dev_read(struct o2hb_region *reg,
1105 				    char *page)
1106 {
1107 	unsigned int ret = 0;
1108 
1109 	if (reg->hr_bdev)
1110 		ret = sprintf(page, "%s\n", reg->hr_dev_name);
1111 
1112 	return ret;
1113 }
1114 
1115 static void o2hb_init_region_params(struct o2hb_region *reg)
1116 {
1117 	reg->hr_slots_per_page = PAGE_CACHE_SIZE >> reg->hr_block_bits;
1118 	reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
1119 
1120 	mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
1121 	     reg->hr_start_block, reg->hr_blocks);
1122 	mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
1123 	     reg->hr_block_bytes, reg->hr_block_bits);
1124 	mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
1125 	mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
1126 }
1127 
1128 static int o2hb_map_slot_data(struct o2hb_region *reg)
1129 {
1130 	int i, j;
1131 	unsigned int last_slot;
1132 	unsigned int spp = reg->hr_slots_per_page;
1133 	struct page *page;
1134 	char *raw;
1135 	struct o2hb_disk_slot *slot;
1136 
1137 	reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
1138 	if (reg->hr_tmp_block == NULL) {
1139 		mlog_errno(-ENOMEM);
1140 		return -ENOMEM;
1141 	}
1142 
1143 	reg->hr_slots = kcalloc(reg->hr_blocks,
1144 				sizeof(struct o2hb_disk_slot), GFP_KERNEL);
1145 	if (reg->hr_slots == NULL) {
1146 		mlog_errno(-ENOMEM);
1147 		return -ENOMEM;
1148 	}
1149 
1150 	for(i = 0; i < reg->hr_blocks; i++) {
1151 		slot = &reg->hr_slots[i];
1152 		slot->ds_node_num = i;
1153 		INIT_LIST_HEAD(&slot->ds_live_item);
1154 		slot->ds_raw_block = NULL;
1155 	}
1156 
1157 	reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
1158 	mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
1159 			   "at %u blocks per page\n",
1160 	     reg->hr_num_pages, reg->hr_blocks, spp);
1161 
1162 	reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
1163 				    GFP_KERNEL);
1164 	if (!reg->hr_slot_data) {
1165 		mlog_errno(-ENOMEM);
1166 		return -ENOMEM;
1167 	}
1168 
1169 	for(i = 0; i < reg->hr_num_pages; i++) {
1170 		page = alloc_page(GFP_KERNEL);
1171 		if (!page) {
1172 			mlog_errno(-ENOMEM);
1173 			return -ENOMEM;
1174 		}
1175 
1176 		reg->hr_slot_data[i] = page;
1177 
1178 		last_slot = i * spp;
1179 		raw = page_address(page);
1180 		for (j = 0;
1181 		     (j < spp) && ((j + last_slot) < reg->hr_blocks);
1182 		     j++) {
1183 			BUG_ON((j + last_slot) >= reg->hr_blocks);
1184 
1185 			slot = &reg->hr_slots[j + last_slot];
1186 			slot->ds_raw_block =
1187 				(struct o2hb_disk_heartbeat_block *) raw;
1188 
1189 			raw += reg->hr_block_bytes;
1190 		}
1191 	}
1192 
1193 	return 0;
1194 }
1195 
1196 /* Read in all the slots available and populate the tracking
1197  * structures so that we can start with a baseline idea of what's
1198  * there. */
1199 static int o2hb_populate_slot_data(struct o2hb_region *reg)
1200 {
1201 	int ret, i;
1202 	struct o2hb_disk_slot *slot;
1203 	struct o2hb_disk_heartbeat_block *hb_block;
1204 
1205 	mlog_entry_void();
1206 
1207 	ret = o2hb_read_slots(reg, reg->hr_blocks);
1208 	if (ret) {
1209 		mlog_errno(ret);
1210 		goto out;
1211 	}
1212 
1213 	/* We only want to get an idea of the values initially in each
1214 	 * slot, so we do no verification - o2hb_check_slot will
1215 	 * actually determine if each configured slot is valid and
1216 	 * whether any values have changed. */
1217 	for(i = 0; i < reg->hr_blocks; i++) {
1218 		slot = &reg->hr_slots[i];
1219 		hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
1220 
1221 		/* Only fill the values that o2hb_check_slot uses to
1222 		 * determine changing slots */
1223 		slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
1224 		slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
1225 	}
1226 
1227 out:
1228 	mlog_exit(ret);
1229 	return ret;
1230 }
1231 
1232 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
1233 static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1234 				     const char *page,
1235 				     size_t count)
1236 {
1237 	struct task_struct *hb_task;
1238 	long fd;
1239 	int sectsize;
1240 	char *p = (char *)page;
1241 	struct file *filp = NULL;
1242 	struct inode *inode = NULL;
1243 	ssize_t ret = -EINVAL;
1244 
1245 	if (reg->hr_bdev)
1246 		goto out;
1247 
1248 	/* We can't heartbeat without having had our node number
1249 	 * configured yet. */
1250 	if (o2nm_this_node() == O2NM_MAX_NODES)
1251 		goto out;
1252 
1253 	fd = simple_strtol(p, &p, 0);
1254 	if (!p || (*p && (*p != '\n')))
1255 		goto out;
1256 
1257 	if (fd < 0 || fd >= INT_MAX)
1258 		goto out;
1259 
1260 	filp = fget(fd);
1261 	if (filp == NULL)
1262 		goto out;
1263 
1264 	if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
1265 	    reg->hr_block_bytes == 0)
1266 		goto out;
1267 
1268 	inode = igrab(filp->f_mapping->host);
1269 	if (inode == NULL)
1270 		goto out;
1271 
1272 	if (!S_ISBLK(inode->i_mode))
1273 		goto out;
1274 
1275 	reg->hr_bdev = I_BDEV(filp->f_mapping->host);
1276 	ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, 0);
1277 	if (ret) {
1278 		reg->hr_bdev = NULL;
1279 		goto out;
1280 	}
1281 	inode = NULL;
1282 
1283 	bdevname(reg->hr_bdev, reg->hr_dev_name);
1284 
1285 	sectsize = bdev_hardsect_size(reg->hr_bdev);
1286 	if (sectsize != reg->hr_block_bytes) {
1287 		mlog(ML_ERROR,
1288 		     "blocksize %u incorrect for device, expected %d",
1289 		     reg->hr_block_bytes, sectsize);
1290 		ret = -EINVAL;
1291 		goto out;
1292 	}
1293 
1294 	o2hb_init_region_params(reg);
1295 
1296 	/* Generation of zero is invalid */
1297 	do {
1298 		get_random_bytes(&reg->hr_generation,
1299 				 sizeof(reg->hr_generation));
1300 	} while (reg->hr_generation == 0);
1301 
1302 	ret = o2hb_map_slot_data(reg);
1303 	if (ret) {
1304 		mlog_errno(ret);
1305 		goto out;
1306 	}
1307 
1308 	ret = o2hb_populate_slot_data(reg);
1309 	if (ret) {
1310 		mlog_errno(ret);
1311 		goto out;
1312 	}
1313 
1314 	INIT_DELAYED_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout);
1315 
1316 	/*
1317 	 * A node is considered live after it has beat LIVE_THRESHOLD
1318 	 * times.  We're not steady until we've given them a chance
1319 	 * _after_ our first read.
1320 	 */
1321 	atomic_set(&reg->hr_steady_iterations, O2HB_LIVE_THRESHOLD + 1);
1322 
1323 	hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
1324 			      reg->hr_item.ci_name);
1325 	if (IS_ERR(hb_task)) {
1326 		ret = PTR_ERR(hb_task);
1327 		mlog_errno(ret);
1328 		goto out;
1329 	}
1330 
1331 	spin_lock(&o2hb_live_lock);
1332 	reg->hr_task = hb_task;
1333 	spin_unlock(&o2hb_live_lock);
1334 
1335 	ret = wait_event_interruptible(o2hb_steady_queue,
1336 				atomic_read(&reg->hr_steady_iterations) == 0);
1337 	if (ret) {
1338 		spin_lock(&o2hb_live_lock);
1339 		hb_task = reg->hr_task;
1340 		reg->hr_task = NULL;
1341 		spin_unlock(&o2hb_live_lock);
1342 
1343 		if (hb_task)
1344 			kthread_stop(hb_task);
1345 		goto out;
1346 	}
1347 
1348 	ret = count;
1349 out:
1350 	if (filp)
1351 		fput(filp);
1352 	if (inode)
1353 		iput(inode);
1354 	if (ret < 0) {
1355 		if (reg->hr_bdev) {
1356 			blkdev_put(reg->hr_bdev);
1357 			reg->hr_bdev = NULL;
1358 		}
1359 	}
1360 	return ret;
1361 }
1362 
1363 static ssize_t o2hb_region_pid_read(struct o2hb_region *reg,
1364                                       char *page)
1365 {
1366 	pid_t pid = 0;
1367 
1368 	spin_lock(&o2hb_live_lock);
1369 	if (reg->hr_task)
1370 		pid = reg->hr_task->pid;
1371 	spin_unlock(&o2hb_live_lock);
1372 
1373 	if (!pid)
1374 		return 0;
1375 
1376 	return sprintf(page, "%u\n", pid);
1377 }
1378 
1379 struct o2hb_region_attribute {
1380 	struct configfs_attribute attr;
1381 	ssize_t (*show)(struct o2hb_region *, char *);
1382 	ssize_t (*store)(struct o2hb_region *, const char *, size_t);
1383 };
1384 
1385 static struct o2hb_region_attribute o2hb_region_attr_block_bytes = {
1386 	.attr	= { .ca_owner = THIS_MODULE,
1387 		    .ca_name = "block_bytes",
1388 		    .ca_mode = S_IRUGO | S_IWUSR },
1389 	.show	= o2hb_region_block_bytes_read,
1390 	.store	= o2hb_region_block_bytes_write,
1391 };
1392 
1393 static struct o2hb_region_attribute o2hb_region_attr_start_block = {
1394 	.attr	= { .ca_owner = THIS_MODULE,
1395 		    .ca_name = "start_block",
1396 		    .ca_mode = S_IRUGO | S_IWUSR },
1397 	.show	= o2hb_region_start_block_read,
1398 	.store	= o2hb_region_start_block_write,
1399 };
1400 
1401 static struct o2hb_region_attribute o2hb_region_attr_blocks = {
1402 	.attr	= { .ca_owner = THIS_MODULE,
1403 		    .ca_name = "blocks",
1404 		    .ca_mode = S_IRUGO | S_IWUSR },
1405 	.show	= o2hb_region_blocks_read,
1406 	.store	= o2hb_region_blocks_write,
1407 };
1408 
1409 static struct o2hb_region_attribute o2hb_region_attr_dev = {
1410 	.attr	= { .ca_owner = THIS_MODULE,
1411 		    .ca_name = "dev",
1412 		    .ca_mode = S_IRUGO | S_IWUSR },
1413 	.show	= o2hb_region_dev_read,
1414 	.store	= o2hb_region_dev_write,
1415 };
1416 
1417 static struct o2hb_region_attribute o2hb_region_attr_pid = {
1418        .attr   = { .ca_owner = THIS_MODULE,
1419                    .ca_name = "pid",
1420                    .ca_mode = S_IRUGO | S_IRUSR },
1421        .show   = o2hb_region_pid_read,
1422 };
1423 
1424 static struct configfs_attribute *o2hb_region_attrs[] = {
1425 	&o2hb_region_attr_block_bytes.attr,
1426 	&o2hb_region_attr_start_block.attr,
1427 	&o2hb_region_attr_blocks.attr,
1428 	&o2hb_region_attr_dev.attr,
1429 	&o2hb_region_attr_pid.attr,
1430 	NULL,
1431 };
1432 
1433 static ssize_t o2hb_region_show(struct config_item *item,
1434 				struct configfs_attribute *attr,
1435 				char *page)
1436 {
1437 	struct o2hb_region *reg = to_o2hb_region(item);
1438 	struct o2hb_region_attribute *o2hb_region_attr =
1439 		container_of(attr, struct o2hb_region_attribute, attr);
1440 	ssize_t ret = 0;
1441 
1442 	if (o2hb_region_attr->show)
1443 		ret = o2hb_region_attr->show(reg, page);
1444 	return ret;
1445 }
1446 
1447 static ssize_t o2hb_region_store(struct config_item *item,
1448 				 struct configfs_attribute *attr,
1449 				 const char *page, size_t count)
1450 {
1451 	struct o2hb_region *reg = to_o2hb_region(item);
1452 	struct o2hb_region_attribute *o2hb_region_attr =
1453 		container_of(attr, struct o2hb_region_attribute, attr);
1454 	ssize_t ret = -EINVAL;
1455 
1456 	if (o2hb_region_attr->store)
1457 		ret = o2hb_region_attr->store(reg, page, count);
1458 	return ret;
1459 }
1460 
1461 static struct configfs_item_operations o2hb_region_item_ops = {
1462 	.release		= o2hb_region_release,
1463 	.show_attribute		= o2hb_region_show,
1464 	.store_attribute	= o2hb_region_store,
1465 };
1466 
1467 static struct config_item_type o2hb_region_type = {
1468 	.ct_item_ops	= &o2hb_region_item_ops,
1469 	.ct_attrs	= o2hb_region_attrs,
1470 	.ct_owner	= THIS_MODULE,
1471 };
1472 
1473 /* heartbeat set */
1474 
1475 struct o2hb_heartbeat_group {
1476 	struct config_group hs_group;
1477 	/* some stuff? */
1478 };
1479 
1480 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
1481 {
1482 	return group ?
1483 		container_of(group, struct o2hb_heartbeat_group, hs_group)
1484 		: NULL;
1485 }
1486 
1487 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
1488 							  const char *name)
1489 {
1490 	struct o2hb_region *reg = NULL;
1491 	struct config_item *ret = NULL;
1492 
1493 	reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
1494 	if (reg == NULL)
1495 		goto out; /* ENOMEM */
1496 
1497 	config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
1498 
1499 	ret = &reg->hr_item;
1500 
1501 	spin_lock(&o2hb_live_lock);
1502 	list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
1503 	spin_unlock(&o2hb_live_lock);
1504 out:
1505 	if (ret == NULL)
1506 		kfree(reg);
1507 
1508 	return ret;
1509 }
1510 
1511 static void o2hb_heartbeat_group_drop_item(struct config_group *group,
1512 					   struct config_item *item)
1513 {
1514 	struct task_struct *hb_task;
1515 	struct o2hb_region *reg = to_o2hb_region(item);
1516 
1517 	/* stop the thread when the user removes the region dir */
1518 	spin_lock(&o2hb_live_lock);
1519 	hb_task = reg->hr_task;
1520 	reg->hr_task = NULL;
1521 	spin_unlock(&o2hb_live_lock);
1522 
1523 	if (hb_task)
1524 		kthread_stop(hb_task);
1525 
1526 	config_item_put(item);
1527 }
1528 
1529 struct o2hb_heartbeat_group_attribute {
1530 	struct configfs_attribute attr;
1531 	ssize_t (*show)(struct o2hb_heartbeat_group *, char *);
1532 	ssize_t (*store)(struct o2hb_heartbeat_group *, const char *, size_t);
1533 };
1534 
1535 static ssize_t o2hb_heartbeat_group_show(struct config_item *item,
1536 					 struct configfs_attribute *attr,
1537 					 char *page)
1538 {
1539 	struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
1540 	struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
1541 		container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
1542 	ssize_t ret = 0;
1543 
1544 	if (o2hb_heartbeat_group_attr->show)
1545 		ret = o2hb_heartbeat_group_attr->show(reg, page);
1546 	return ret;
1547 }
1548 
1549 static ssize_t o2hb_heartbeat_group_store(struct config_item *item,
1550 					  struct configfs_attribute *attr,
1551 					  const char *page, size_t count)
1552 {
1553 	struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
1554 	struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
1555 		container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
1556 	ssize_t ret = -EINVAL;
1557 
1558 	if (o2hb_heartbeat_group_attr->store)
1559 		ret = o2hb_heartbeat_group_attr->store(reg, page, count);
1560 	return ret;
1561 }
1562 
1563 static ssize_t o2hb_heartbeat_group_threshold_show(struct o2hb_heartbeat_group *group,
1564 						     char *page)
1565 {
1566 	return sprintf(page, "%u\n", o2hb_dead_threshold);
1567 }
1568 
1569 static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group *group,
1570 						    const char *page,
1571 						    size_t count)
1572 {
1573 	unsigned long tmp;
1574 	char *p = (char *)page;
1575 
1576 	tmp = simple_strtoul(p, &p, 10);
1577 	if (!p || (*p && (*p != '\n')))
1578                 return -EINVAL;
1579 
1580 	/* this will validate ranges for us. */
1581 	o2hb_dead_threshold_set((unsigned int) tmp);
1582 
1583 	return count;
1584 }
1585 
1586 static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = {
1587 	.attr	= { .ca_owner = THIS_MODULE,
1588 		    .ca_name = "dead_threshold",
1589 		    .ca_mode = S_IRUGO | S_IWUSR },
1590 	.show	= o2hb_heartbeat_group_threshold_show,
1591 	.store	= o2hb_heartbeat_group_threshold_store,
1592 };
1593 
1594 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
1595 	&o2hb_heartbeat_group_attr_threshold.attr,
1596 	NULL,
1597 };
1598 
1599 static struct configfs_item_operations o2hb_hearbeat_group_item_ops = {
1600 	.show_attribute		= o2hb_heartbeat_group_show,
1601 	.store_attribute	= o2hb_heartbeat_group_store,
1602 };
1603 
1604 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
1605 	.make_item	= o2hb_heartbeat_group_make_item,
1606 	.drop_item	= o2hb_heartbeat_group_drop_item,
1607 };
1608 
1609 static struct config_item_type o2hb_heartbeat_group_type = {
1610 	.ct_group_ops	= &o2hb_heartbeat_group_group_ops,
1611 	.ct_item_ops	= &o2hb_hearbeat_group_item_ops,
1612 	.ct_attrs	= o2hb_heartbeat_group_attrs,
1613 	.ct_owner	= THIS_MODULE,
1614 };
1615 
1616 /* this is just here to avoid touching group in heartbeat.h which the
1617  * entire damn world #includes */
1618 struct config_group *o2hb_alloc_hb_set(void)
1619 {
1620 	struct o2hb_heartbeat_group *hs = NULL;
1621 	struct config_group *ret = NULL;
1622 
1623 	hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
1624 	if (hs == NULL)
1625 		goto out;
1626 
1627 	config_group_init_type_name(&hs->hs_group, "heartbeat",
1628 				    &o2hb_heartbeat_group_type);
1629 
1630 	ret = &hs->hs_group;
1631 out:
1632 	if (ret == NULL)
1633 		kfree(hs);
1634 	return ret;
1635 }
1636 
1637 void o2hb_free_hb_set(struct config_group *group)
1638 {
1639 	struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
1640 	kfree(hs);
1641 }
1642 
1643 /* hb callback registration and issueing */
1644 
1645 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
1646 {
1647 	if (type == O2HB_NUM_CB)
1648 		return ERR_PTR(-EINVAL);
1649 
1650 	return &o2hb_callbacks[type];
1651 }
1652 
1653 void o2hb_setup_callback(struct o2hb_callback_func *hc,
1654 			 enum o2hb_callback_type type,
1655 			 o2hb_cb_func *func,
1656 			 void *data,
1657 			 int priority)
1658 {
1659 	INIT_LIST_HEAD(&hc->hc_item);
1660 	hc->hc_func = func;
1661 	hc->hc_data = data;
1662 	hc->hc_priority = priority;
1663 	hc->hc_type = type;
1664 	hc->hc_magic = O2HB_CB_MAGIC;
1665 }
1666 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
1667 
1668 int o2hb_register_callback(struct o2hb_callback_func *hc)
1669 {
1670 	struct o2hb_callback_func *tmp;
1671 	struct list_head *iter;
1672 	struct o2hb_callback *hbcall;
1673 	int ret;
1674 
1675 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
1676 	BUG_ON(!list_empty(&hc->hc_item));
1677 
1678 	hbcall = hbcall_from_type(hc->hc_type);
1679 	if (IS_ERR(hbcall)) {
1680 		ret = PTR_ERR(hbcall);
1681 		goto out;
1682 	}
1683 
1684 	down_write(&o2hb_callback_sem);
1685 
1686 	list_for_each(iter, &hbcall->list) {
1687 		tmp = list_entry(iter, struct o2hb_callback_func, hc_item);
1688 		if (hc->hc_priority < tmp->hc_priority) {
1689 			list_add_tail(&hc->hc_item, iter);
1690 			break;
1691 		}
1692 	}
1693 	if (list_empty(&hc->hc_item))
1694 		list_add_tail(&hc->hc_item, &hbcall->list);
1695 
1696 	up_write(&o2hb_callback_sem);
1697 	ret = 0;
1698 out:
1699 	mlog(ML_HEARTBEAT, "returning %d on behalf of %p for funcs %p\n",
1700 	     ret, __builtin_return_address(0), hc);
1701 	return ret;
1702 }
1703 EXPORT_SYMBOL_GPL(o2hb_register_callback);
1704 
1705 void o2hb_unregister_callback(struct o2hb_callback_func *hc)
1706 {
1707 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
1708 
1709 	mlog(ML_HEARTBEAT, "on behalf of %p for funcs %p\n",
1710 	     __builtin_return_address(0), hc);
1711 
1712 	if (list_empty(&hc->hc_item))
1713 		return;
1714 
1715 	down_write(&o2hb_callback_sem);
1716 
1717 	list_del_init(&hc->hc_item);
1718 
1719 	up_write(&o2hb_callback_sem);
1720 }
1721 EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
1722 
1723 int o2hb_check_node_heartbeating(u8 node_num)
1724 {
1725 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1726 
1727 	o2hb_fill_node_map(testing_map, sizeof(testing_map));
1728 	if (!test_bit(node_num, testing_map)) {
1729 		mlog(ML_HEARTBEAT,
1730 		     "node (%u) does not have heartbeating enabled.\n",
1731 		     node_num);
1732 		return 0;
1733 	}
1734 
1735 	return 1;
1736 }
1737 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating);
1738 
1739 int o2hb_check_node_heartbeating_from_callback(u8 node_num)
1740 {
1741 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1742 
1743 	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
1744 	if (!test_bit(node_num, testing_map)) {
1745 		mlog(ML_HEARTBEAT,
1746 		     "node (%u) does not have heartbeating enabled.\n",
1747 		     node_num);
1748 		return 0;
1749 	}
1750 
1751 	return 1;
1752 }
1753 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
1754 
1755 /* Makes sure our local node is configured with a node number, and is
1756  * heartbeating. */
1757 int o2hb_check_local_node_heartbeating(void)
1758 {
1759 	u8 node_num;
1760 
1761 	/* if this node was set then we have networking */
1762 	node_num = o2nm_this_node();
1763 	if (node_num == O2NM_MAX_NODES) {
1764 		mlog(ML_HEARTBEAT, "this node has not been configured.\n");
1765 		return 0;
1766 	}
1767 
1768 	return o2hb_check_node_heartbeating(node_num);
1769 }
1770 EXPORT_SYMBOL_GPL(o2hb_check_local_node_heartbeating);
1771 
1772 /*
1773  * this is just a hack until we get the plumbing which flips file systems
1774  * read only and drops the hb ref instead of killing the node dead.
1775  */
1776 void o2hb_stop_all_regions(void)
1777 {
1778 	struct o2hb_region *reg;
1779 
1780 	mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
1781 
1782 	spin_lock(&o2hb_live_lock);
1783 
1784 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
1785 		reg->hr_unclean_stop = 1;
1786 
1787 	spin_unlock(&o2hb_live_lock);
1788 }
1789 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
1790