xref: /openbmc/linux/fs/ocfs2/cluster/heartbeat.c (revision 87c2ce3b)
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 spinlock_t o2hb_live_lock = SPIN_LOCK_UNLOCKED;
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 work_struct	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 };
157 
158 static void o2hb_write_timeout(void *arg)
159 {
160 	struct o2hb_region *reg = arg;
161 
162 	mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
163 	     "milliseconds\n", reg->hr_dev_name,
164 	     jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
165 	o2quo_disk_timeout();
166 }
167 
168 static void o2hb_arm_write_timeout(struct o2hb_region *reg)
169 {
170 	mlog(0, "Queue write timeout for %u ms\n", O2HB_MAX_WRITE_TIMEOUT_MS);
171 
172 	cancel_delayed_work(&reg->hr_write_timeout_work);
173 	reg->hr_last_timeout_start = jiffies;
174 	schedule_delayed_work(&reg->hr_write_timeout_work,
175 			      msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
176 }
177 
178 static void o2hb_disarm_write_timeout(struct o2hb_region *reg)
179 {
180 	cancel_delayed_work(&reg->hr_write_timeout_work);
181 	flush_scheduled_work();
182 }
183 
184 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc,
185 				      unsigned int num_ios)
186 {
187 	atomic_set(&wc->wc_num_reqs, num_ios);
188 	init_completion(&wc->wc_io_complete);
189 }
190 
191 /* Used in error paths too */
192 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
193 				     unsigned int num)
194 {
195 	/* sadly atomic_sub_and_test() isn't available on all platforms.  The
196 	 * good news is that the fast path only completes one at a time */
197 	while(num--) {
198 		if (atomic_dec_and_test(&wc->wc_num_reqs)) {
199 			BUG_ON(num > 0);
200 			complete(&wc->wc_io_complete);
201 		}
202 	}
203 }
204 
205 static void o2hb_wait_on_io(struct o2hb_region *reg,
206 			    struct o2hb_bio_wait_ctxt *wc)
207 {
208 	struct address_space *mapping = reg->hr_bdev->bd_inode->i_mapping;
209 
210 	blk_run_address_space(mapping);
211 
212 	wait_for_completion(&wc->wc_io_complete);
213 }
214 
215 static int o2hb_bio_end_io(struct bio *bio,
216 			   unsigned int bytes_done,
217 			   int error)
218 {
219 	struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
220 
221 	if (error)
222 		mlog(ML_ERROR, "IO Error %d\n", error);
223 
224 	if (bio->bi_size)
225 		return 1;
226 
227 	o2hb_bio_wait_dec(wc, 1);
228 	return 0;
229 }
230 
231 /* Setup a Bio to cover I/O against num_slots slots starting at
232  * start_slot. */
233 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
234 				      struct o2hb_bio_wait_ctxt *wc,
235 				      unsigned int start_slot,
236 				      unsigned int num_slots)
237 {
238 	int i, nr_vecs, len, first_page, last_page;
239 	unsigned int vec_len, vec_start;
240 	unsigned int bits = reg->hr_block_bits;
241 	unsigned int spp = reg->hr_slots_per_page;
242 	struct bio *bio;
243 	struct page *page;
244 
245 	nr_vecs = (num_slots + spp - 1) / spp;
246 
247 	/* Testing has shown this allocation to take long enough under
248 	 * GFP_KERNEL that the local node can get fenced. It would be
249 	 * nicest if we could pre-allocate these bios and avoid this
250 	 * all together. */
251 	bio = bio_alloc(GFP_ATOMIC, nr_vecs);
252 	if (!bio) {
253 		mlog(ML_ERROR, "Could not alloc slots BIO!\n");
254 		bio = ERR_PTR(-ENOMEM);
255 		goto bail;
256 	}
257 
258 	/* Must put everything in 512 byte sectors for the bio... */
259 	bio->bi_sector = (reg->hr_start_block + start_slot) << (bits - 9);
260 	bio->bi_bdev = reg->hr_bdev;
261 	bio->bi_private = wc;
262 	bio->bi_end_io = o2hb_bio_end_io;
263 
264 	first_page = start_slot / spp;
265 	last_page = first_page + nr_vecs;
266 	vec_start = (start_slot << bits) % PAGE_CACHE_SIZE;
267 	for(i = first_page; i < last_page; i++) {
268 		page = reg->hr_slot_data[i];
269 
270 		vec_len = PAGE_CACHE_SIZE;
271 		/* last page might be short */
272 		if (((i + 1) * spp) > (start_slot + num_slots))
273 			vec_len = ((num_slots + start_slot) % spp) << bits;
274 		vec_len -=  vec_start;
275 
276 		mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
277 		     i, vec_len, vec_start);
278 
279 		len = bio_add_page(bio, page, vec_len, vec_start);
280 		if (len != vec_len) {
281 			bio_put(bio);
282 			bio = ERR_PTR(-EIO);
283 
284 			mlog(ML_ERROR, "Error adding page to bio i = %d, "
285 			     "vec_len = %u, len = %d\n, start = %u\n",
286 			     i, vec_len, len, vec_start);
287 			goto bail;
288 		}
289 
290 		vec_start = 0;
291 	}
292 
293 bail:
294 	return bio;
295 }
296 
297 /*
298  * Compute the maximum number of sectors the bdev can handle in one bio,
299  * as a power of two.
300  *
301  * Stolen from oracleasm, thanks Joel!
302  */
303 static int compute_max_sectors(struct block_device *bdev)
304 {
305 	int max_pages, max_sectors, pow_two_sectors;
306 
307 	struct request_queue *q;
308 
309 	q = bdev_get_queue(bdev);
310 	max_pages = q->max_sectors >> (PAGE_SHIFT - 9);
311 	if (max_pages > BIO_MAX_PAGES)
312 		max_pages = BIO_MAX_PAGES;
313 	if (max_pages > q->max_phys_segments)
314 		max_pages = q->max_phys_segments;
315 	if (max_pages > q->max_hw_segments)
316 		max_pages = q->max_hw_segments;
317 	max_pages--; /* Handle I/Os that straddle a page */
318 
319 	max_sectors = max_pages << (PAGE_SHIFT - 9);
320 
321 	/* Why is fls() 1-based???? */
322 	pow_two_sectors = 1 << (fls(max_sectors) - 1);
323 
324 	return pow_two_sectors;
325 }
326 
327 static inline void o2hb_compute_request_limits(struct o2hb_region *reg,
328 					       unsigned int num_slots,
329 					       unsigned int *num_bios,
330 					       unsigned int *slots_per_bio)
331 {
332 	unsigned int max_sectors, io_sectors;
333 
334 	max_sectors = compute_max_sectors(reg->hr_bdev);
335 
336 	io_sectors = num_slots << (reg->hr_block_bits - 9);
337 
338 	*num_bios = (io_sectors + max_sectors - 1) / max_sectors;
339 	*slots_per_bio = max_sectors >> (reg->hr_block_bits - 9);
340 
341 	mlog(ML_HB_BIO, "My io size is %u sectors for %u slots. This "
342 	     "device can handle %u sectors of I/O\n", io_sectors, num_slots,
343 	     max_sectors);
344 	mlog(ML_HB_BIO, "Will need %u bios holding %u slots each\n",
345 	     *num_bios, *slots_per_bio);
346 }
347 
348 static int o2hb_read_slots(struct o2hb_region *reg,
349 			   unsigned int max_slots)
350 {
351 	unsigned int num_bios, slots_per_bio, start_slot, num_slots;
352 	int i, status;
353 	struct o2hb_bio_wait_ctxt wc;
354 	struct bio **bios;
355 	struct bio *bio;
356 
357 	o2hb_compute_request_limits(reg, max_slots, &num_bios, &slots_per_bio);
358 
359 	bios = kcalloc(num_bios, sizeof(struct bio *), GFP_KERNEL);
360 	if (!bios) {
361 		status = -ENOMEM;
362 		mlog_errno(status);
363 		return status;
364 	}
365 
366 	o2hb_bio_wait_init(&wc, num_bios);
367 
368 	num_slots = slots_per_bio;
369 	for(i = 0; i < num_bios; i++) {
370 		start_slot = i * slots_per_bio;
371 
372 		/* adjust num_slots at last bio */
373 		if (max_slots < (start_slot + num_slots))
374 			num_slots = max_slots - start_slot;
375 
376 		bio = o2hb_setup_one_bio(reg, &wc, start_slot, num_slots);
377 		if (IS_ERR(bio)) {
378 			o2hb_bio_wait_dec(&wc, num_bios - i);
379 
380 			status = PTR_ERR(bio);
381 			mlog_errno(status);
382 			goto bail_and_wait;
383 		}
384 		bios[i] = bio;
385 
386 		submit_bio(READ, bio);
387 	}
388 
389 	status = 0;
390 
391 bail_and_wait:
392 	o2hb_wait_on_io(reg, &wc);
393 
394 	if (bios) {
395 		for(i = 0; i < num_bios; i++)
396 			if (bios[i])
397 				bio_put(bios[i]);
398 		kfree(bios);
399 	}
400 
401 	return status;
402 }
403 
404 static int o2hb_issue_node_write(struct o2hb_region *reg,
405 				 struct bio **write_bio,
406 				 struct o2hb_bio_wait_ctxt *write_wc)
407 {
408 	int status;
409 	unsigned int slot;
410 	struct bio *bio;
411 
412 	o2hb_bio_wait_init(write_wc, 1);
413 
414 	slot = o2nm_this_node();
415 
416 	bio = o2hb_setup_one_bio(reg, write_wc, slot, 1);
417 	if (IS_ERR(bio)) {
418 		status = PTR_ERR(bio);
419 		mlog_errno(status);
420 		goto bail;
421 	}
422 
423 	submit_bio(WRITE, bio);
424 
425 	*write_bio = bio;
426 	status = 0;
427 bail:
428 	return status;
429 }
430 
431 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
432 				     struct o2hb_disk_heartbeat_block *hb_block)
433 {
434 	__le32 old_cksum;
435 	u32 ret;
436 
437 	/* We want to compute the block crc with a 0 value in the
438 	 * hb_cksum field. Save it off here and replace after the
439 	 * crc. */
440 	old_cksum = hb_block->hb_cksum;
441 	hb_block->hb_cksum = 0;
442 
443 	ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
444 
445 	hb_block->hb_cksum = old_cksum;
446 
447 	return ret;
448 }
449 
450 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
451 {
452 	mlog(ML_ERROR, "Dump slot information: seq = 0x%"MLFx64", node = %u, "
453 	     "cksum = 0x%x, generation 0x%"MLFx64"\n",
454 	     le64_to_cpu(hb_block->hb_seq), hb_block->hb_node,
455 	     le32_to_cpu(hb_block->hb_cksum),
456 	     le64_to_cpu(hb_block->hb_generation));
457 }
458 
459 static int o2hb_verify_crc(struct o2hb_region *reg,
460 			   struct o2hb_disk_heartbeat_block *hb_block)
461 {
462 	u32 read, computed;
463 
464 	read = le32_to_cpu(hb_block->hb_cksum);
465 	computed = o2hb_compute_block_crc_le(reg, hb_block);
466 
467 	return read == computed;
468 }
469 
470 /* We want to make sure that nobody is heartbeating on top of us --
471  * this will help detect an invalid configuration. */
472 static int o2hb_check_last_timestamp(struct o2hb_region *reg)
473 {
474 	int node_num, ret;
475 	struct o2hb_disk_slot *slot;
476 	struct o2hb_disk_heartbeat_block *hb_block;
477 
478 	node_num = o2nm_this_node();
479 
480 	ret = 1;
481 	slot = &reg->hr_slots[node_num];
482 	/* Don't check on our 1st timestamp */
483 	if (slot->ds_last_time) {
484 		hb_block = slot->ds_raw_block;
485 
486 		if (le64_to_cpu(hb_block->hb_seq) != slot->ds_last_time)
487 			ret = 0;
488 	}
489 
490 	return ret;
491 }
492 
493 static inline void o2hb_prepare_block(struct o2hb_region *reg,
494 				      u64 generation)
495 {
496 	int node_num;
497 	u64 cputime;
498 	struct o2hb_disk_slot *slot;
499 	struct o2hb_disk_heartbeat_block *hb_block;
500 
501 	node_num = o2nm_this_node();
502 	slot = &reg->hr_slots[node_num];
503 
504 	hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
505 	memset(hb_block, 0, reg->hr_block_bytes);
506 	/* TODO: time stuff */
507 	cputime = CURRENT_TIME.tv_sec;
508 	if (!cputime)
509 		cputime = 1;
510 
511 	hb_block->hb_seq = cpu_to_le64(cputime);
512 	hb_block->hb_node = node_num;
513 	hb_block->hb_generation = cpu_to_le64(generation);
514 
515 	/* This step must always happen last! */
516 	hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
517 								   hb_block));
518 
519 	mlog(ML_HB_BIO, "our node generation = 0x%"MLFx64", cksum = 0x%x\n",
520 	     cpu_to_le64(generation), le32_to_cpu(hb_block->hb_cksum));
521 }
522 
523 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
524 				struct o2nm_node *node,
525 				int idx)
526 {
527 	struct list_head *iter;
528 	struct o2hb_callback_func *f;
529 
530 	list_for_each(iter, &hbcall->list) {
531 		f = list_entry(iter, struct o2hb_callback_func, hc_item);
532 		mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
533 		(f->hc_func)(node, idx, f->hc_data);
534 	}
535 }
536 
537 /* Will run the list in order until we process the passed event */
538 static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
539 {
540 	int empty;
541 	struct o2hb_callback *hbcall;
542 	struct o2hb_node_event *event;
543 
544 	spin_lock(&o2hb_live_lock);
545 	empty = list_empty(&queued_event->hn_item);
546 	spin_unlock(&o2hb_live_lock);
547 	if (empty)
548 		return;
549 
550 	/* Holding callback sem assures we don't alter the callback
551 	 * lists when doing this, and serializes ourselves with other
552 	 * processes wanting callbacks. */
553 	down_write(&o2hb_callback_sem);
554 
555 	spin_lock(&o2hb_live_lock);
556 	while (!list_empty(&o2hb_node_events)
557 	       && !list_empty(&queued_event->hn_item)) {
558 		event = list_entry(o2hb_node_events.next,
559 				   struct o2hb_node_event,
560 				   hn_item);
561 		list_del_init(&event->hn_item);
562 		spin_unlock(&o2hb_live_lock);
563 
564 		mlog(ML_HEARTBEAT, "Node %s event for %d\n",
565 		     event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
566 		     event->hn_node_num);
567 
568 		hbcall = hbcall_from_type(event->hn_event_type);
569 
570 		/* We should *never* have gotten on to the list with a
571 		 * bad type... This isn't something that we should try
572 		 * to recover from. */
573 		BUG_ON(IS_ERR(hbcall));
574 
575 		o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
576 
577 		spin_lock(&o2hb_live_lock);
578 	}
579 	spin_unlock(&o2hb_live_lock);
580 
581 	up_write(&o2hb_callback_sem);
582 }
583 
584 static void o2hb_queue_node_event(struct o2hb_node_event *event,
585 				  enum o2hb_callback_type type,
586 				  struct o2nm_node *node,
587 				  int node_num)
588 {
589 	assert_spin_locked(&o2hb_live_lock);
590 
591 	event->hn_event_type = type;
592 	event->hn_node = node;
593 	event->hn_node_num = node_num;
594 
595 	mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
596 	     type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
597 
598 	list_add_tail(&event->hn_item, &o2hb_node_events);
599 }
600 
601 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
602 {
603 	struct o2hb_node_event event =
604 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
605 	struct o2nm_node *node;
606 
607 	node = o2nm_get_node_by_num(slot->ds_node_num);
608 	if (!node)
609 		return;
610 
611 	spin_lock(&o2hb_live_lock);
612 	if (!list_empty(&slot->ds_live_item)) {
613 		mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
614 		     slot->ds_node_num);
615 
616 		list_del_init(&slot->ds_live_item);
617 
618 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
619 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
620 
621 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
622 					      slot->ds_node_num);
623 		}
624 	}
625 	spin_unlock(&o2hb_live_lock);
626 
627 	o2hb_run_event_list(&event);
628 
629 	o2nm_node_put(node);
630 }
631 
632 static int o2hb_check_slot(struct o2hb_region *reg,
633 			   struct o2hb_disk_slot *slot)
634 {
635 	int changed = 0, gen_changed = 0;
636 	struct o2hb_node_event event =
637 		{ .hn_item = LIST_HEAD_INIT(event.hn_item), };
638 	struct o2nm_node *node;
639 	struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
640 	u64 cputime;
641 
642 	memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
643 
644 	/* Is this correct? Do we assume that the node doesn't exist
645 	 * if we're not configured for him? */
646 	node = o2nm_get_node_by_num(slot->ds_node_num);
647 	if (!node)
648 		return 0;
649 
650 	if (!o2hb_verify_crc(reg, hb_block)) {
651 		/* all paths from here will drop o2hb_live_lock for
652 		 * us. */
653 		spin_lock(&o2hb_live_lock);
654 
655 		/* Don't print an error on the console in this case -
656 		 * a freshly formatted heartbeat area will not have a
657 		 * crc set on it. */
658 		if (list_empty(&slot->ds_live_item))
659 			goto out;
660 
661 		/* The node is live but pushed out a bad crc. We
662 		 * consider it a transient miss but don't populate any
663 		 * other values as they may be junk. */
664 		mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
665 		     slot->ds_node_num, reg->hr_dev_name);
666 		o2hb_dump_slot(hb_block);
667 
668 		slot->ds_equal_samples++;
669 		goto fire_callbacks;
670 	}
671 
672 	/* we don't care if these wrap.. the state transitions below
673 	 * clear at the right places */
674 	cputime = le64_to_cpu(hb_block->hb_seq);
675 	if (slot->ds_last_time != cputime)
676 		slot->ds_changed_samples++;
677 	else
678 		slot->ds_equal_samples++;
679 	slot->ds_last_time = cputime;
680 
681 	/* The node changed heartbeat generations. We assume this to
682 	 * mean it dropped off but came back before we timed out. We
683 	 * want to consider it down for the time being but don't want
684 	 * to lose any changed_samples state we might build up to
685 	 * considering it live again. */
686 	if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
687 		gen_changed = 1;
688 		slot->ds_equal_samples = 0;
689 		mlog(ML_HEARTBEAT, "Node %d changed generation (0x%"MLFx64" "
690 		     "to 0x%"MLFx64")\n", slot->ds_node_num,
691 		     slot->ds_last_generation,
692 		     le64_to_cpu(hb_block->hb_generation));
693 	}
694 
695 	slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
696 
697 	mlog(ML_HEARTBEAT, "Slot %d gen 0x%"MLFx64" cksum 0x%x "
698 	     "seq %"MLFu64" last %"MLFu64" changed %u equal %u\n",
699 	     slot->ds_node_num, slot->ds_last_generation,
700 	     le32_to_cpu(hb_block->hb_cksum), le64_to_cpu(hb_block->hb_seq),
701 	     slot->ds_last_time, slot->ds_changed_samples,
702 	     slot->ds_equal_samples);
703 
704 	spin_lock(&o2hb_live_lock);
705 
706 fire_callbacks:
707 	/* dead nodes only come to life after some number of
708 	 * changes at any time during their dead time */
709 	if (list_empty(&slot->ds_live_item) &&
710 	    slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
711 		mlog(ML_HEARTBEAT, "Node %d (id 0x%"MLFx64") joined my "
712 		     "region\n", slot->ds_node_num, slot->ds_last_generation);
713 
714 		/* first on the list generates a callback */
715 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
716 			set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
717 
718 			o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
719 					      slot->ds_node_num);
720 
721 			changed = 1;
722 		}
723 
724 		list_add_tail(&slot->ds_live_item,
725 			      &o2hb_live_slots[slot->ds_node_num]);
726 
727 		slot->ds_equal_samples = 0;
728 		goto out;
729 	}
730 
731 	/* if the list is dead, we're done.. */
732 	if (list_empty(&slot->ds_live_item))
733 		goto out;
734 
735 	/* live nodes only go dead after enough consequtive missed
736 	 * samples..  reset the missed counter whenever we see
737 	 * activity */
738 	if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
739 		mlog(ML_HEARTBEAT, "Node %d left my region\n",
740 		     slot->ds_node_num);
741 
742 		/* last off the live_slot generates a callback */
743 		list_del_init(&slot->ds_live_item);
744 		if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
745 			clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
746 
747 			o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
748 					      slot->ds_node_num);
749 
750 			changed = 1;
751 		}
752 
753 		/* We don't clear this because the node is still
754 		 * actually writing new blocks. */
755 		if (!gen_changed)
756 			slot->ds_changed_samples = 0;
757 		goto out;
758 	}
759 	if (slot->ds_changed_samples) {
760 		slot->ds_changed_samples = 0;
761 		slot->ds_equal_samples = 0;
762 	}
763 out:
764 	spin_unlock(&o2hb_live_lock);
765 
766 	o2hb_run_event_list(&event);
767 
768 	o2nm_node_put(node);
769 	return changed;
770 }
771 
772 /* This could be faster if we just implmented a find_last_bit, but I
773  * don't think the circumstances warrant it. */
774 static int o2hb_highest_node(unsigned long *nodes,
775 			     int numbits)
776 {
777 	int highest, node;
778 
779 	highest = numbits;
780 	node = -1;
781 	while ((node = find_next_bit(nodes, numbits, node + 1)) != -1) {
782 		if (node >= numbits)
783 			break;
784 
785 		highest = node;
786 	}
787 
788 	return highest;
789 }
790 
791 static void o2hb_do_disk_heartbeat(struct o2hb_region *reg)
792 {
793 	int i, ret, highest_node, change = 0;
794 	unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
795 	struct bio *write_bio;
796 	struct o2hb_bio_wait_ctxt write_wc;
797 
798 	if (o2nm_configured_node_map(configured_nodes, sizeof(configured_nodes)))
799 		return;
800 
801 	highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
802 	if (highest_node >= O2NM_MAX_NODES) {
803 		mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n");
804 		return;
805 	}
806 
807 	/* No sense in reading the slots of nodes that don't exist
808 	 * yet. Of course, if the node definitions have holes in them
809 	 * then we're reading an empty slot anyway... Consider this
810 	 * best-effort. */
811 	ret = o2hb_read_slots(reg, highest_node + 1);
812 	if (ret < 0) {
813 		mlog_errno(ret);
814 		return;
815 	}
816 
817 	/* With an up to date view of the slots, we can check that no
818 	 * other node has been improperly configured to heartbeat in
819 	 * our slot. */
820 	if (!o2hb_check_last_timestamp(reg))
821 		mlog(ML_ERROR, "Device \"%s\": another node is heartbeating "
822 		     "in our slot!\n", reg->hr_dev_name);
823 
824 	/* fill in the proper info for our next heartbeat */
825 	o2hb_prepare_block(reg, reg->hr_generation);
826 
827 	/* And fire off the write. Note that we don't wait on this I/O
828 	 * until later. */
829 	ret = o2hb_issue_node_write(reg, &write_bio, &write_wc);
830 	if (ret < 0) {
831 		mlog_errno(ret);
832 		return;
833 	}
834 
835 	i = -1;
836 	while((i = find_next_bit(configured_nodes, O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
837 
838 		change |= o2hb_check_slot(reg, &reg->hr_slots[i]);
839 	}
840 
841 	/*
842 	 * We have to be sure we've advertised ourselves on disk
843 	 * before we can go to steady state.  This ensures that
844 	 * people we find in our steady state have seen us.
845 	 */
846 	o2hb_wait_on_io(reg, &write_wc);
847 	bio_put(write_bio);
848 	o2hb_arm_write_timeout(reg);
849 
850 	/* let the person who launched us know when things are steady */
851 	if (!change && (atomic_read(&reg->hr_steady_iterations) != 0)) {
852 		if (atomic_dec_and_test(&reg->hr_steady_iterations))
853 			wake_up(&o2hb_steady_queue);
854 	}
855 }
856 
857 /* Subtract b from a, storing the result in a. a *must* have a larger
858  * value than b. */
859 static void o2hb_tv_subtract(struct timeval *a,
860 			     struct timeval *b)
861 {
862 	/* just return 0 when a is after b */
863 	if (a->tv_sec < b->tv_sec ||
864 	    (a->tv_sec == b->tv_sec && a->tv_usec < b->tv_usec)) {
865 		a->tv_sec = 0;
866 		a->tv_usec = 0;
867 		return;
868 	}
869 
870 	a->tv_sec -= b->tv_sec;
871 	a->tv_usec -= b->tv_usec;
872 	while ( a->tv_usec < 0 ) {
873 		a->tv_sec--;
874 		a->tv_usec += 1000000;
875 	}
876 }
877 
878 static unsigned int o2hb_elapsed_msecs(struct timeval *start,
879 				       struct timeval *end)
880 {
881 	struct timeval res = *end;
882 
883 	o2hb_tv_subtract(&res, start);
884 
885 	return res.tv_sec * 1000 + res.tv_usec / 1000;
886 }
887 
888 /*
889  * we ride the region ref that the region dir holds.  before the region
890  * dir is removed and drops it ref it will wait to tear down this
891  * thread.
892  */
893 static int o2hb_thread(void *data)
894 {
895 	int i, ret;
896 	struct o2hb_region *reg = data;
897 	struct bio *write_bio;
898 	struct o2hb_bio_wait_ctxt write_wc;
899 	struct timeval before_hb, after_hb;
900 	unsigned int elapsed_msec;
901 
902 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
903 
904 	set_user_nice(current, -20);
905 
906 	while (!kthread_should_stop() && !reg->hr_unclean_stop) {
907 		/* We track the time spent inside
908 		 * o2hb_do_disk_heartbeat so that we avoid more then
909 		 * hr_timeout_ms between disk writes. On busy systems
910 		 * this should result in a heartbeat which is less
911 		 * likely to time itself out. */
912 		do_gettimeofday(&before_hb);
913 
914 		o2hb_do_disk_heartbeat(reg);
915 
916 		do_gettimeofday(&after_hb);
917 		elapsed_msec = o2hb_elapsed_msecs(&before_hb, &after_hb);
918 
919 		mlog(0, "start = %lu.%lu, end = %lu.%lu, msec = %u\n",
920 		     before_hb.tv_sec, before_hb.tv_usec,
921 		     after_hb.tv_sec, after_hb.tv_usec, elapsed_msec);
922 
923 		if (elapsed_msec < reg->hr_timeout_ms) {
924 			/* the kthread api has blocked signals for us so no
925 			 * need to record the return value. */
926 			msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
927 		}
928 	}
929 
930 	o2hb_disarm_write_timeout(reg);
931 
932 	/* unclean stop is only used in very bad situation */
933 	for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
934 		o2hb_shutdown_slot(&reg->hr_slots[i]);
935 
936 	/* Explicit down notification - avoid forcing the other nodes
937 	 * to timeout on this region when we could just as easily
938 	 * write a clear generation - thus indicating to them that
939 	 * this node has left this region.
940 	 *
941 	 * XXX: Should we skip this on unclean_stop? */
942 	o2hb_prepare_block(reg, 0);
943 	ret = o2hb_issue_node_write(reg, &write_bio, &write_wc);
944 	if (ret == 0) {
945 		o2hb_wait_on_io(reg, &write_wc);
946 		bio_put(write_bio);
947 	} else {
948 		mlog_errno(ret);
949 	}
950 
951 	mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n");
952 
953 	return 0;
954 }
955 
956 void o2hb_init(void)
957 {
958 	int i;
959 
960 	for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
961 		INIT_LIST_HEAD(&o2hb_callbacks[i].list);
962 
963 	for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
964 		INIT_LIST_HEAD(&o2hb_live_slots[i]);
965 
966 	INIT_LIST_HEAD(&o2hb_node_events);
967 
968 	memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
969 }
970 
971 /* if we're already in a callback then we're already serialized by the sem */
972 static void o2hb_fill_node_map_from_callback(unsigned long *map,
973 					     unsigned bytes)
974 {
975 	BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
976 
977 	memcpy(map, &o2hb_live_node_bitmap, bytes);
978 }
979 
980 /*
981  * get a map of all nodes that are heartbeating in any regions
982  */
983 void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
984 {
985 	/* callers want to serialize this map and callbacks so that they
986 	 * can trust that they don't miss nodes coming to the party */
987 	down_read(&o2hb_callback_sem);
988 	spin_lock(&o2hb_live_lock);
989 	o2hb_fill_node_map_from_callback(map, bytes);
990 	spin_unlock(&o2hb_live_lock);
991 	up_read(&o2hb_callback_sem);
992 }
993 EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
994 
995 /*
996  * heartbeat configfs bits.  The heartbeat set is a default set under
997  * the cluster set in nodemanager.c.
998  */
999 
1000 static struct o2hb_region *to_o2hb_region(struct config_item *item)
1001 {
1002 	return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
1003 }
1004 
1005 /* drop_item only drops its ref after killing the thread, nothing should
1006  * be using the region anymore.  this has to clean up any state that
1007  * attributes might have built up. */
1008 static void o2hb_region_release(struct config_item *item)
1009 {
1010 	int i;
1011 	struct page *page;
1012 	struct o2hb_region *reg = to_o2hb_region(item);
1013 
1014 	if (reg->hr_tmp_block)
1015 		kfree(reg->hr_tmp_block);
1016 
1017 	if (reg->hr_slot_data) {
1018 		for (i = 0; i < reg->hr_num_pages; i++) {
1019 			page = reg->hr_slot_data[i];
1020 			if (page)
1021 				__free_page(page);
1022 		}
1023 		kfree(reg->hr_slot_data);
1024 	}
1025 
1026 	if (reg->hr_bdev)
1027 		blkdev_put(reg->hr_bdev);
1028 
1029 	if (reg->hr_slots)
1030 		kfree(reg->hr_slots);
1031 
1032 	spin_lock(&o2hb_live_lock);
1033 	list_del(&reg->hr_all_item);
1034 	spin_unlock(&o2hb_live_lock);
1035 
1036 	kfree(reg);
1037 }
1038 
1039 static int o2hb_read_block_input(struct o2hb_region *reg,
1040 				 const char *page,
1041 				 size_t count,
1042 				 unsigned long *ret_bytes,
1043 				 unsigned int *ret_bits)
1044 {
1045 	unsigned long bytes;
1046 	char *p = (char *)page;
1047 
1048 	bytes = simple_strtoul(p, &p, 0);
1049 	if (!p || (*p && (*p != '\n')))
1050 		return -EINVAL;
1051 
1052 	/* Heartbeat and fs min / max block sizes are the same. */
1053 	if (bytes > 4096 || bytes < 512)
1054 		return -ERANGE;
1055 	if (hweight16(bytes) != 1)
1056 		return -EINVAL;
1057 
1058 	if (ret_bytes)
1059 		*ret_bytes = bytes;
1060 	if (ret_bits)
1061 		*ret_bits = ffs(bytes) - 1;
1062 
1063 	return 0;
1064 }
1065 
1066 static ssize_t o2hb_region_block_bytes_read(struct o2hb_region *reg,
1067 					    char *page)
1068 {
1069 	return sprintf(page, "%u\n", reg->hr_block_bytes);
1070 }
1071 
1072 static ssize_t o2hb_region_block_bytes_write(struct o2hb_region *reg,
1073 					     const char *page,
1074 					     size_t count)
1075 {
1076 	int status;
1077 	unsigned long block_bytes;
1078 	unsigned int block_bits;
1079 
1080 	if (reg->hr_bdev)
1081 		return -EINVAL;
1082 
1083 	status = o2hb_read_block_input(reg, page, count,
1084 				       &block_bytes, &block_bits);
1085 	if (status)
1086 		return status;
1087 
1088 	reg->hr_block_bytes = (unsigned int)block_bytes;
1089 	reg->hr_block_bits = block_bits;
1090 
1091 	return count;
1092 }
1093 
1094 static ssize_t o2hb_region_start_block_read(struct o2hb_region *reg,
1095 					    char *page)
1096 {
1097 	return sprintf(page, "%llu\n", reg->hr_start_block);
1098 }
1099 
1100 static ssize_t o2hb_region_start_block_write(struct o2hb_region *reg,
1101 					     const char *page,
1102 					     size_t count)
1103 {
1104 	unsigned long long tmp;
1105 	char *p = (char *)page;
1106 
1107 	if (reg->hr_bdev)
1108 		return -EINVAL;
1109 
1110 	tmp = simple_strtoull(p, &p, 0);
1111 	if (!p || (*p && (*p != '\n')))
1112 		return -EINVAL;
1113 
1114 	reg->hr_start_block = tmp;
1115 
1116 	return count;
1117 }
1118 
1119 static ssize_t o2hb_region_blocks_read(struct o2hb_region *reg,
1120 				       char *page)
1121 {
1122 	return sprintf(page, "%d\n", reg->hr_blocks);
1123 }
1124 
1125 static ssize_t o2hb_region_blocks_write(struct o2hb_region *reg,
1126 					const char *page,
1127 					size_t count)
1128 {
1129 	unsigned long tmp;
1130 	char *p = (char *)page;
1131 
1132 	if (reg->hr_bdev)
1133 		return -EINVAL;
1134 
1135 	tmp = simple_strtoul(p, &p, 0);
1136 	if (!p || (*p && (*p != '\n')))
1137 		return -EINVAL;
1138 
1139 	if (tmp > O2NM_MAX_NODES || tmp == 0)
1140 		return -ERANGE;
1141 
1142 	reg->hr_blocks = (unsigned int)tmp;
1143 
1144 	return count;
1145 }
1146 
1147 static ssize_t o2hb_region_dev_read(struct o2hb_region *reg,
1148 				    char *page)
1149 {
1150 	unsigned int ret = 0;
1151 
1152 	if (reg->hr_bdev)
1153 		ret = sprintf(page, "%s\n", reg->hr_dev_name);
1154 
1155 	return ret;
1156 }
1157 
1158 static void o2hb_init_region_params(struct o2hb_region *reg)
1159 {
1160 	reg->hr_slots_per_page = PAGE_CACHE_SIZE >> reg->hr_block_bits;
1161 	reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
1162 
1163 	mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
1164 	     reg->hr_start_block, reg->hr_blocks);
1165 	mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
1166 	     reg->hr_block_bytes, reg->hr_block_bits);
1167 	mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
1168 	mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
1169 }
1170 
1171 static int o2hb_map_slot_data(struct o2hb_region *reg)
1172 {
1173 	int i, j;
1174 	unsigned int last_slot;
1175 	unsigned int spp = reg->hr_slots_per_page;
1176 	struct page *page;
1177 	char *raw;
1178 	struct o2hb_disk_slot *slot;
1179 
1180 	reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
1181 	if (reg->hr_tmp_block == NULL) {
1182 		mlog_errno(-ENOMEM);
1183 		return -ENOMEM;
1184 	}
1185 
1186 	reg->hr_slots = kcalloc(reg->hr_blocks,
1187 				sizeof(struct o2hb_disk_slot), GFP_KERNEL);
1188 	if (reg->hr_slots == NULL) {
1189 		mlog_errno(-ENOMEM);
1190 		return -ENOMEM;
1191 	}
1192 
1193 	for(i = 0; i < reg->hr_blocks; i++) {
1194 		slot = &reg->hr_slots[i];
1195 		slot->ds_node_num = i;
1196 		INIT_LIST_HEAD(&slot->ds_live_item);
1197 		slot->ds_raw_block = NULL;
1198 	}
1199 
1200 	reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
1201 	mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
1202 			   "at %u blocks per page\n",
1203 	     reg->hr_num_pages, reg->hr_blocks, spp);
1204 
1205 	reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
1206 				    GFP_KERNEL);
1207 	if (!reg->hr_slot_data) {
1208 		mlog_errno(-ENOMEM);
1209 		return -ENOMEM;
1210 	}
1211 
1212 	for(i = 0; i < reg->hr_num_pages; i++) {
1213 		page = alloc_page(GFP_KERNEL);
1214 		if (!page) {
1215 			mlog_errno(-ENOMEM);
1216 			return -ENOMEM;
1217 		}
1218 
1219 		reg->hr_slot_data[i] = page;
1220 
1221 		last_slot = i * spp;
1222 		raw = page_address(page);
1223 		for (j = 0;
1224 		     (j < spp) && ((j + last_slot) < reg->hr_blocks);
1225 		     j++) {
1226 			BUG_ON((j + last_slot) >= reg->hr_blocks);
1227 
1228 			slot = &reg->hr_slots[j + last_slot];
1229 			slot->ds_raw_block =
1230 				(struct o2hb_disk_heartbeat_block *) raw;
1231 
1232 			raw += reg->hr_block_bytes;
1233 		}
1234 	}
1235 
1236 	return 0;
1237 }
1238 
1239 /* Read in all the slots available and populate the tracking
1240  * structures so that we can start with a baseline idea of what's
1241  * there. */
1242 static int o2hb_populate_slot_data(struct o2hb_region *reg)
1243 {
1244 	int ret, i;
1245 	struct o2hb_disk_slot *slot;
1246 	struct o2hb_disk_heartbeat_block *hb_block;
1247 
1248 	mlog_entry_void();
1249 
1250 	ret = o2hb_read_slots(reg, reg->hr_blocks);
1251 	if (ret) {
1252 		mlog_errno(ret);
1253 		goto out;
1254 	}
1255 
1256 	/* We only want to get an idea of the values initially in each
1257 	 * slot, so we do no verification - o2hb_check_slot will
1258 	 * actually determine if each configured slot is valid and
1259 	 * whether any values have changed. */
1260 	for(i = 0; i < reg->hr_blocks; i++) {
1261 		slot = &reg->hr_slots[i];
1262 		hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
1263 
1264 		/* Only fill the values that o2hb_check_slot uses to
1265 		 * determine changing slots */
1266 		slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
1267 		slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
1268 	}
1269 
1270 out:
1271 	mlog_exit(ret);
1272 	return ret;
1273 }
1274 
1275 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
1276 static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1277 				     const char *page,
1278 				     size_t count)
1279 {
1280 	long fd;
1281 	int sectsize;
1282 	char *p = (char *)page;
1283 	struct file *filp = NULL;
1284 	struct inode *inode = NULL;
1285 	ssize_t ret = -EINVAL;
1286 
1287 	if (reg->hr_bdev)
1288 		goto out;
1289 
1290 	/* We can't heartbeat without having had our node number
1291 	 * configured yet. */
1292 	if (o2nm_this_node() == O2NM_MAX_NODES)
1293 		goto out;
1294 
1295 	fd = simple_strtol(p, &p, 0);
1296 	if (!p || (*p && (*p != '\n')))
1297 		goto out;
1298 
1299 	if (fd < 0 || fd >= INT_MAX)
1300 		goto out;
1301 
1302 	filp = fget(fd);
1303 	if (filp == NULL)
1304 		goto out;
1305 
1306 	if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
1307 	    reg->hr_block_bytes == 0)
1308 		goto out;
1309 
1310 	inode = igrab(filp->f_mapping->host);
1311 	if (inode == NULL)
1312 		goto out;
1313 
1314 	if (!S_ISBLK(inode->i_mode))
1315 		goto out;
1316 
1317 	reg->hr_bdev = I_BDEV(filp->f_mapping->host);
1318 	ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, 0);
1319 	if (ret) {
1320 		reg->hr_bdev = NULL;
1321 		goto out;
1322 	}
1323 	inode = NULL;
1324 
1325 	bdevname(reg->hr_bdev, reg->hr_dev_name);
1326 
1327 	sectsize = bdev_hardsect_size(reg->hr_bdev);
1328 	if (sectsize != reg->hr_block_bytes) {
1329 		mlog(ML_ERROR,
1330 		     "blocksize %u incorrect for device, expected %d",
1331 		     reg->hr_block_bytes, sectsize);
1332 		ret = -EINVAL;
1333 		goto out;
1334 	}
1335 
1336 	o2hb_init_region_params(reg);
1337 
1338 	/* Generation of zero is invalid */
1339 	do {
1340 		get_random_bytes(&reg->hr_generation,
1341 				 sizeof(reg->hr_generation));
1342 	} while (reg->hr_generation == 0);
1343 
1344 	ret = o2hb_map_slot_data(reg);
1345 	if (ret) {
1346 		mlog_errno(ret);
1347 		goto out;
1348 	}
1349 
1350 	ret = o2hb_populate_slot_data(reg);
1351 	if (ret) {
1352 		mlog_errno(ret);
1353 		goto out;
1354 	}
1355 
1356 	INIT_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout, reg);
1357 
1358 	/*
1359 	 * A node is considered live after it has beat LIVE_THRESHOLD
1360 	 * times.  We're not steady until we've given them a chance
1361 	 * _after_ our first read.
1362 	 */
1363 	atomic_set(&reg->hr_steady_iterations, O2HB_LIVE_THRESHOLD + 1);
1364 
1365 	reg->hr_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
1366 				   reg->hr_item.ci_name);
1367 	if (IS_ERR(reg->hr_task)) {
1368 		ret = PTR_ERR(reg->hr_task);
1369 		mlog_errno(ret);
1370 		reg->hr_task = NULL;
1371 		goto out;
1372 	}
1373 
1374 	ret = wait_event_interruptible(o2hb_steady_queue,
1375 				atomic_read(&reg->hr_steady_iterations) == 0);
1376 	if (ret) {
1377 		kthread_stop(reg->hr_task);
1378 		reg->hr_task = NULL;
1379 		goto out;
1380 	}
1381 
1382 	ret = count;
1383 out:
1384 	if (filp)
1385 		fput(filp);
1386 	if (inode)
1387 		iput(inode);
1388 	if (ret < 0) {
1389 		if (reg->hr_bdev) {
1390 			blkdev_put(reg->hr_bdev);
1391 			reg->hr_bdev = NULL;
1392 		}
1393 	}
1394 	return ret;
1395 }
1396 
1397 struct o2hb_region_attribute {
1398 	struct configfs_attribute attr;
1399 	ssize_t (*show)(struct o2hb_region *, char *);
1400 	ssize_t (*store)(struct o2hb_region *, const char *, size_t);
1401 };
1402 
1403 static struct o2hb_region_attribute o2hb_region_attr_block_bytes = {
1404 	.attr	= { .ca_owner = THIS_MODULE,
1405 		    .ca_name = "block_bytes",
1406 		    .ca_mode = S_IRUGO | S_IWUSR },
1407 	.show	= o2hb_region_block_bytes_read,
1408 	.store	= o2hb_region_block_bytes_write,
1409 };
1410 
1411 static struct o2hb_region_attribute o2hb_region_attr_start_block = {
1412 	.attr	= { .ca_owner = THIS_MODULE,
1413 		    .ca_name = "start_block",
1414 		    .ca_mode = S_IRUGO | S_IWUSR },
1415 	.show	= o2hb_region_start_block_read,
1416 	.store	= o2hb_region_start_block_write,
1417 };
1418 
1419 static struct o2hb_region_attribute o2hb_region_attr_blocks = {
1420 	.attr	= { .ca_owner = THIS_MODULE,
1421 		    .ca_name = "blocks",
1422 		    .ca_mode = S_IRUGO | S_IWUSR },
1423 	.show	= o2hb_region_blocks_read,
1424 	.store	= o2hb_region_blocks_write,
1425 };
1426 
1427 static struct o2hb_region_attribute o2hb_region_attr_dev = {
1428 	.attr	= { .ca_owner = THIS_MODULE,
1429 		    .ca_name = "dev",
1430 		    .ca_mode = S_IRUGO | S_IWUSR },
1431 	.show	= o2hb_region_dev_read,
1432 	.store	= o2hb_region_dev_write,
1433 };
1434 
1435 static struct configfs_attribute *o2hb_region_attrs[] = {
1436 	&o2hb_region_attr_block_bytes.attr,
1437 	&o2hb_region_attr_start_block.attr,
1438 	&o2hb_region_attr_blocks.attr,
1439 	&o2hb_region_attr_dev.attr,
1440 	NULL,
1441 };
1442 
1443 static ssize_t o2hb_region_show(struct config_item *item,
1444 				struct configfs_attribute *attr,
1445 				char *page)
1446 {
1447 	struct o2hb_region *reg = to_o2hb_region(item);
1448 	struct o2hb_region_attribute *o2hb_region_attr =
1449 		container_of(attr, struct o2hb_region_attribute, attr);
1450 	ssize_t ret = 0;
1451 
1452 	if (o2hb_region_attr->show)
1453 		ret = o2hb_region_attr->show(reg, page);
1454 	return ret;
1455 }
1456 
1457 static ssize_t o2hb_region_store(struct config_item *item,
1458 				 struct configfs_attribute *attr,
1459 				 const char *page, size_t count)
1460 {
1461 	struct o2hb_region *reg = to_o2hb_region(item);
1462 	struct o2hb_region_attribute *o2hb_region_attr =
1463 		container_of(attr, struct o2hb_region_attribute, attr);
1464 	ssize_t ret = -EINVAL;
1465 
1466 	if (o2hb_region_attr->store)
1467 		ret = o2hb_region_attr->store(reg, page, count);
1468 	return ret;
1469 }
1470 
1471 static struct configfs_item_operations o2hb_region_item_ops = {
1472 	.release		= o2hb_region_release,
1473 	.show_attribute		= o2hb_region_show,
1474 	.store_attribute	= o2hb_region_store,
1475 };
1476 
1477 static struct config_item_type o2hb_region_type = {
1478 	.ct_item_ops	= &o2hb_region_item_ops,
1479 	.ct_attrs	= o2hb_region_attrs,
1480 	.ct_owner	= THIS_MODULE,
1481 };
1482 
1483 /* heartbeat set */
1484 
1485 struct o2hb_heartbeat_group {
1486 	struct config_group hs_group;
1487 	/* some stuff? */
1488 };
1489 
1490 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
1491 {
1492 	return group ?
1493 		container_of(group, struct o2hb_heartbeat_group, hs_group)
1494 		: NULL;
1495 }
1496 
1497 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
1498 							  const char *name)
1499 {
1500 	struct o2hb_region *reg = NULL;
1501 	struct config_item *ret = NULL;
1502 
1503 	reg = kcalloc(1, sizeof(struct o2hb_region), GFP_KERNEL);
1504 	if (reg == NULL)
1505 		goto out; /* ENOMEM */
1506 
1507 	config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
1508 
1509 	ret = &reg->hr_item;
1510 
1511 	spin_lock(&o2hb_live_lock);
1512 	list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
1513 	spin_unlock(&o2hb_live_lock);
1514 out:
1515 	if (ret == NULL)
1516 		kfree(reg);
1517 
1518 	return ret;
1519 }
1520 
1521 static void o2hb_heartbeat_group_drop_item(struct config_group *group,
1522 					   struct config_item *item)
1523 {
1524 	struct o2hb_region *reg = to_o2hb_region(item);
1525 
1526 	/* stop the thread when the user removes the region dir */
1527 	if (reg->hr_task) {
1528 		kthread_stop(reg->hr_task);
1529 		reg->hr_task = NULL;
1530 	}
1531 
1532 	config_item_put(item);
1533 }
1534 
1535 struct o2hb_heartbeat_group_attribute {
1536 	struct configfs_attribute attr;
1537 	ssize_t (*show)(struct o2hb_heartbeat_group *, char *);
1538 	ssize_t (*store)(struct o2hb_heartbeat_group *, const char *, size_t);
1539 };
1540 
1541 static ssize_t o2hb_heartbeat_group_show(struct config_item *item,
1542 					 struct configfs_attribute *attr,
1543 					 char *page)
1544 {
1545 	struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
1546 	struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
1547 		container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
1548 	ssize_t ret = 0;
1549 
1550 	if (o2hb_heartbeat_group_attr->show)
1551 		ret = o2hb_heartbeat_group_attr->show(reg, page);
1552 	return ret;
1553 }
1554 
1555 static ssize_t o2hb_heartbeat_group_store(struct config_item *item,
1556 					  struct configfs_attribute *attr,
1557 					  const char *page, size_t count)
1558 {
1559 	struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
1560 	struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
1561 		container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
1562 	ssize_t ret = -EINVAL;
1563 
1564 	if (o2hb_heartbeat_group_attr->store)
1565 		ret = o2hb_heartbeat_group_attr->store(reg, page, count);
1566 	return ret;
1567 }
1568 
1569 static ssize_t o2hb_heartbeat_group_threshold_show(struct o2hb_heartbeat_group *group,
1570 						     char *page)
1571 {
1572 	return sprintf(page, "%u\n", o2hb_dead_threshold);
1573 }
1574 
1575 static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group *group,
1576 						    const char *page,
1577 						    size_t count)
1578 {
1579 	unsigned long tmp;
1580 	char *p = (char *)page;
1581 
1582 	tmp = simple_strtoul(p, &p, 10);
1583 	if (!p || (*p && (*p != '\n')))
1584                 return -EINVAL;
1585 
1586 	/* this will validate ranges for us. */
1587 	o2hb_dead_threshold_set((unsigned int) tmp);
1588 
1589 	return count;
1590 }
1591 
1592 static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = {
1593 	.attr	= { .ca_owner = THIS_MODULE,
1594 		    .ca_name = "dead_threshold",
1595 		    .ca_mode = S_IRUGO | S_IWUSR },
1596 	.show	= o2hb_heartbeat_group_threshold_show,
1597 	.store	= o2hb_heartbeat_group_threshold_store,
1598 };
1599 
1600 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
1601 	&o2hb_heartbeat_group_attr_threshold.attr,
1602 	NULL,
1603 };
1604 
1605 static struct configfs_item_operations o2hb_hearbeat_group_item_ops = {
1606 	.show_attribute		= o2hb_heartbeat_group_show,
1607 	.store_attribute	= o2hb_heartbeat_group_store,
1608 };
1609 
1610 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
1611 	.make_item	= o2hb_heartbeat_group_make_item,
1612 	.drop_item	= o2hb_heartbeat_group_drop_item,
1613 };
1614 
1615 static struct config_item_type o2hb_heartbeat_group_type = {
1616 	.ct_group_ops	= &o2hb_heartbeat_group_group_ops,
1617 	.ct_item_ops	= &o2hb_hearbeat_group_item_ops,
1618 	.ct_attrs	= o2hb_heartbeat_group_attrs,
1619 	.ct_owner	= THIS_MODULE,
1620 };
1621 
1622 /* this is just here to avoid touching group in heartbeat.h which the
1623  * entire damn world #includes */
1624 struct config_group *o2hb_alloc_hb_set(void)
1625 {
1626 	struct o2hb_heartbeat_group *hs = NULL;
1627 	struct config_group *ret = NULL;
1628 
1629 	hs = kcalloc(1, sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
1630 	if (hs == NULL)
1631 		goto out;
1632 
1633 	config_group_init_type_name(&hs->hs_group, "heartbeat",
1634 				    &o2hb_heartbeat_group_type);
1635 
1636 	ret = &hs->hs_group;
1637 out:
1638 	if (ret == NULL)
1639 		kfree(hs);
1640 	return ret;
1641 }
1642 
1643 void o2hb_free_hb_set(struct config_group *group)
1644 {
1645 	struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
1646 	kfree(hs);
1647 }
1648 
1649 /* hb callback registration and issueing */
1650 
1651 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
1652 {
1653 	if (type == O2HB_NUM_CB)
1654 		return ERR_PTR(-EINVAL);
1655 
1656 	return &o2hb_callbacks[type];
1657 }
1658 
1659 void o2hb_setup_callback(struct o2hb_callback_func *hc,
1660 			 enum o2hb_callback_type type,
1661 			 o2hb_cb_func *func,
1662 			 void *data,
1663 			 int priority)
1664 {
1665 	INIT_LIST_HEAD(&hc->hc_item);
1666 	hc->hc_func = func;
1667 	hc->hc_data = data;
1668 	hc->hc_priority = priority;
1669 	hc->hc_type = type;
1670 	hc->hc_magic = O2HB_CB_MAGIC;
1671 }
1672 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
1673 
1674 int o2hb_register_callback(struct o2hb_callback_func *hc)
1675 {
1676 	struct o2hb_callback_func *tmp;
1677 	struct list_head *iter;
1678 	struct o2hb_callback *hbcall;
1679 	int ret;
1680 
1681 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
1682 	BUG_ON(!list_empty(&hc->hc_item));
1683 
1684 	hbcall = hbcall_from_type(hc->hc_type);
1685 	if (IS_ERR(hbcall)) {
1686 		ret = PTR_ERR(hbcall);
1687 		goto out;
1688 	}
1689 
1690 	down_write(&o2hb_callback_sem);
1691 
1692 	list_for_each(iter, &hbcall->list) {
1693 		tmp = list_entry(iter, struct o2hb_callback_func, hc_item);
1694 		if (hc->hc_priority < tmp->hc_priority) {
1695 			list_add_tail(&hc->hc_item, iter);
1696 			break;
1697 		}
1698 	}
1699 	if (list_empty(&hc->hc_item))
1700 		list_add_tail(&hc->hc_item, &hbcall->list);
1701 
1702 	up_write(&o2hb_callback_sem);
1703 	ret = 0;
1704 out:
1705 	mlog(ML_HEARTBEAT, "returning %d on behalf of %p for funcs %p\n",
1706 	     ret, __builtin_return_address(0), hc);
1707 	return ret;
1708 }
1709 EXPORT_SYMBOL_GPL(o2hb_register_callback);
1710 
1711 int o2hb_unregister_callback(struct o2hb_callback_func *hc)
1712 {
1713 	BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
1714 
1715 	mlog(ML_HEARTBEAT, "on behalf of %p for funcs %p\n",
1716 	     __builtin_return_address(0), hc);
1717 
1718 	if (list_empty(&hc->hc_item))
1719 		return 0;
1720 
1721 	down_write(&o2hb_callback_sem);
1722 
1723 	list_del_init(&hc->hc_item);
1724 
1725 	up_write(&o2hb_callback_sem);
1726 
1727 	return 0;
1728 }
1729 EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
1730 
1731 int o2hb_check_node_heartbeating(u8 node_num)
1732 {
1733 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1734 
1735 	o2hb_fill_node_map(testing_map, sizeof(testing_map));
1736 	if (!test_bit(node_num, testing_map)) {
1737 		mlog(ML_HEARTBEAT,
1738 		     "node (%u) does not have heartbeating enabled.\n",
1739 		     node_num);
1740 		return 0;
1741 	}
1742 
1743 	return 1;
1744 }
1745 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating);
1746 
1747 int o2hb_check_node_heartbeating_from_callback(u8 node_num)
1748 {
1749 	unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1750 
1751 	o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
1752 	if (!test_bit(node_num, testing_map)) {
1753 		mlog(ML_HEARTBEAT,
1754 		     "node (%u) does not have heartbeating enabled.\n",
1755 		     node_num);
1756 		return 0;
1757 	}
1758 
1759 	return 1;
1760 }
1761 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
1762 
1763 /* Makes sure our local node is configured with a node number, and is
1764  * heartbeating. */
1765 int o2hb_check_local_node_heartbeating(void)
1766 {
1767 	u8 node_num;
1768 
1769 	/* if this node was set then we have networking */
1770 	node_num = o2nm_this_node();
1771 	if (node_num == O2NM_MAX_NODES) {
1772 		mlog(ML_HEARTBEAT, "this node has not been configured.\n");
1773 		return 0;
1774 	}
1775 
1776 	return o2hb_check_node_heartbeating(node_num);
1777 }
1778 EXPORT_SYMBOL_GPL(o2hb_check_local_node_heartbeating);
1779 
1780 /*
1781  * this is just a hack until we get the plumbing which flips file systems
1782  * read only and drops the hb ref instead of killing the node dead.
1783  */
1784 void o2hb_stop_all_regions(void)
1785 {
1786 	struct o2hb_region *reg;
1787 
1788 	mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
1789 
1790 	spin_lock(&o2hb_live_lock);
1791 
1792 	list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
1793 		reg->hr_unclean_stop = 1;
1794 
1795 	spin_unlock(&o2hb_live_lock);
1796 }
1797 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
1798