1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* -*- mode: c; c-basic-offset: 8; -*-
3 * vim: noexpandtab sw=8 ts=8 sts=0:
4 *
5 * Copyright (C) 2004, 2005 Oracle. All rights reserved.
6 */
7
8 #include <linux/kernel.h>
9 #include <linux/sched.h>
10 #include <linux/jiffies.h>
11 #include <linux/module.h>
12 #include <linux/fs.h>
13 #include <linux/bio.h>
14 #include <linux/blkdev.h>
15 #include <linux/delay.h>
16 #include <linux/file.h>
17 #include <linux/kthread.h>
18 #include <linux/configfs.h>
19 #include <linux/random.h>
20 #include <linux/crc32.h>
21 #include <linux/time.h>
22 #include <linux/debugfs.h>
23 #include <linux/slab.h>
24 #include <linux/bitmap.h>
25 #include <linux/ktime.h>
26 #include "heartbeat.h"
27 #include "tcp.h"
28 #include "nodemanager.h"
29 #include "quorum.h"
30
31 #include "masklog.h"
32
33
34 /*
35 * The first heartbeat pass had one global thread that would serialize all hb
36 * callback calls. This global serializing sem should only be removed once
37 * we've made sure that all callees can deal with being called concurrently
38 * from multiple hb region threads.
39 */
40 static DECLARE_RWSEM(o2hb_callback_sem);
41
42 /*
43 * multiple hb threads are watching multiple regions. A node is live
44 * whenever any of the threads sees activity from the node in its region.
45 */
46 static DEFINE_SPINLOCK(o2hb_live_lock);
47 static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
48 static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
49 static LIST_HEAD(o2hb_node_events);
50 static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
51
52 /*
53 * In global heartbeat, we maintain a series of region bitmaps.
54 * - o2hb_region_bitmap allows us to limit the region number to max region.
55 * - o2hb_live_region_bitmap tracks live regions (seen steady iterations).
56 * - o2hb_quorum_region_bitmap tracks live regions that have seen all nodes
57 * heartbeat on it.
58 * - o2hb_failed_region_bitmap tracks the regions that have seen io timeouts.
59 */
60 static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
61 static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
62 static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
63 static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
64
65 #define O2HB_DB_TYPE_LIVENODES 0
66 #define O2HB_DB_TYPE_LIVEREGIONS 1
67 #define O2HB_DB_TYPE_QUORUMREGIONS 2
68 #define O2HB_DB_TYPE_FAILEDREGIONS 3
69 #define O2HB_DB_TYPE_REGION_LIVENODES 4
70 #define O2HB_DB_TYPE_REGION_NUMBER 5
71 #define O2HB_DB_TYPE_REGION_ELAPSED_TIME 6
72 #define O2HB_DB_TYPE_REGION_PINNED 7
73 struct o2hb_debug_buf {
74 int db_type;
75 int db_size;
76 int db_len;
77 void *db_data;
78 };
79
80 static struct o2hb_debug_buf *o2hb_db_livenodes;
81 static struct o2hb_debug_buf *o2hb_db_liveregions;
82 static struct o2hb_debug_buf *o2hb_db_quorumregions;
83 static struct o2hb_debug_buf *o2hb_db_failedregions;
84
85 #define O2HB_DEBUG_DIR "o2hb"
86 #define O2HB_DEBUG_LIVENODES "livenodes"
87 #define O2HB_DEBUG_LIVEREGIONS "live_regions"
88 #define O2HB_DEBUG_QUORUMREGIONS "quorum_regions"
89 #define O2HB_DEBUG_FAILEDREGIONS "failed_regions"
90 #define O2HB_DEBUG_REGION_NUMBER "num"
91 #define O2HB_DEBUG_REGION_ELAPSED_TIME "elapsed_time_in_ms"
92 #define O2HB_DEBUG_REGION_PINNED "pinned"
93
94 static struct dentry *o2hb_debug_dir;
95 static struct dentry *o2hb_debug_livenodes;
96 static struct dentry *o2hb_debug_liveregions;
97 static struct dentry *o2hb_debug_quorumregions;
98 static struct dentry *o2hb_debug_failedregions;
99
100 static LIST_HEAD(o2hb_all_regions);
101
102 static struct o2hb_callback {
103 struct list_head list;
104 } o2hb_callbacks[O2HB_NUM_CB];
105
106 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
107
108 #define O2HB_DEFAULT_BLOCK_BITS 9
109
110 enum o2hb_heartbeat_modes {
111 O2HB_HEARTBEAT_LOCAL = 0,
112 O2HB_HEARTBEAT_GLOBAL,
113 O2HB_HEARTBEAT_NUM_MODES,
114 };
115
116 static const char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = {
117 "local", /* O2HB_HEARTBEAT_LOCAL */
118 "global", /* O2HB_HEARTBEAT_GLOBAL */
119 };
120
121 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
122 static unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL;
123
124 /*
125 * o2hb_dependent_users tracks the number of registered callbacks that depend
126 * on heartbeat. o2net and o2dlm are two entities that register this callback.
127 * However only o2dlm depends on the heartbeat. It does not want the heartbeat
128 * to stop while a dlm domain is still active.
129 */
130 static unsigned int o2hb_dependent_users;
131
132 /*
133 * In global heartbeat mode, all regions are pinned if there are one or more
134 * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All
135 * regions are unpinned if the region count exceeds the cut off or the number
136 * of dependent users falls to zero.
137 */
138 #define O2HB_PIN_CUT_OFF 3
139
140 /*
141 * In local heartbeat mode, we assume the dlm domain name to be the same as
142 * region uuid. This is true for domains created for the file system but not
143 * necessarily true for userdlm domains. This is a known limitation.
144 *
145 * In global heartbeat mode, we pin/unpin all o2hb regions. This solution
146 * works for both file system and userdlm domains.
147 */
148 static int o2hb_region_pin(const char *region_uuid);
149 static void o2hb_region_unpin(const char *region_uuid);
150
151 /* Only sets a new threshold if there are no active regions.
152 *
153 * No locking or otherwise interesting code is required for reading
154 * o2hb_dead_threshold as it can't change once regions are active and
155 * it's not interesting to anyone until then anyway. */
156 static void o2hb_dead_threshold_set(unsigned int threshold)
157 {
158 if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
159 spin_lock(&o2hb_live_lock);
160 if (list_empty(&o2hb_all_regions))
161 o2hb_dead_threshold = threshold;
162 spin_unlock(&o2hb_live_lock);
163 }
164 }
165
166 static int o2hb_global_heartbeat_mode_set(unsigned int hb_mode)
167 {
168 int ret = -1;
169
170 if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) {
171 spin_lock(&o2hb_live_lock);
172 if (list_empty(&o2hb_all_regions)) {
173 o2hb_heartbeat_mode = hb_mode;
174 ret = 0;
175 }
176 spin_unlock(&o2hb_live_lock);
177 }
178
179 return ret;
180 }
181
182 struct o2hb_node_event {
183 struct list_head hn_item;
184 enum o2hb_callback_type hn_event_type;
185 struct o2nm_node *hn_node;
186 int hn_node_num;
187 };
188
189 struct o2hb_disk_slot {
190 struct o2hb_disk_heartbeat_block *ds_raw_block;
191 u8 ds_node_num;
192 u64 ds_last_time;
193 u64 ds_last_generation;
194 u16 ds_equal_samples;
195 u16 ds_changed_samples;
196 struct list_head ds_live_item;
197 };
198
199 /* each thread owns a region.. when we're asked to tear down the region
200 * we ask the thread to stop, who cleans up the region */
201 struct o2hb_region {
202 struct config_item hr_item;
203
204 struct list_head hr_all_item;
205 unsigned hr_unclean_stop:1,
206 hr_aborted_start:1,
207 hr_item_pinned:1,
208 hr_item_dropped:1,
209 hr_node_deleted:1;
210
211 /* protected by the hr_callback_sem */
212 struct task_struct *hr_task;
213
214 unsigned int hr_blocks;
215 unsigned long long hr_start_block;
216
217 unsigned int hr_block_bits;
218 unsigned int hr_block_bytes;
219
220 unsigned int hr_slots_per_page;
221 unsigned int hr_num_pages;
222
223 struct page **hr_slot_data;
224 struct block_device *hr_bdev;
225 struct o2hb_disk_slot *hr_slots;
226
227 /* live node map of this region */
228 unsigned long hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
229 unsigned int hr_region_num;
230
231 struct dentry *hr_debug_dir;
232 struct dentry *hr_debug_livenodes;
233 struct dentry *hr_debug_regnum;
234 struct dentry *hr_debug_elapsed_time;
235 struct dentry *hr_debug_pinned;
236 struct o2hb_debug_buf *hr_db_livenodes;
237 struct o2hb_debug_buf *hr_db_regnum;
238 struct o2hb_debug_buf *hr_db_elapsed_time;
239 struct o2hb_debug_buf *hr_db_pinned;
240
241 /* let the person setting up hb wait for it to return until it
242 * has reached a 'steady' state. This will be fixed when we have
243 * a more complete api that doesn't lead to this sort of fragility. */
244 atomic_t hr_steady_iterations;
245
246 /* terminate o2hb thread if it does not reach steady state
247 * (hr_steady_iterations == 0) within hr_unsteady_iterations */
248 atomic_t hr_unsteady_iterations;
249
250 char hr_dev_name[BDEVNAME_SIZE];
251
252 unsigned int hr_timeout_ms;
253
254 /* randomized as the region goes up and down so that a node
255 * recognizes a node going up and down in one iteration */
256 u64 hr_generation;
257
258 struct delayed_work hr_write_timeout_work;
259 unsigned long hr_last_timeout_start;
260
261 /* negotiate timer, used to negotiate extending hb timeout. */
262 struct delayed_work hr_nego_timeout_work;
263 unsigned long hr_nego_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
264
265 /* Used during o2hb_check_slot to hold a copy of the block
266 * being checked because we temporarily have to zero out the
267 * crc field. */
268 struct o2hb_disk_heartbeat_block *hr_tmp_block;
269
270 /* Message key for negotiate timeout message. */
271 unsigned int hr_key;
272 struct list_head hr_handler_list;
273
274 /* last hb status, 0 for success, other value for error. */
275 int hr_last_hb_status;
276 };
277
278 struct o2hb_bio_wait_ctxt {
279 atomic_t wc_num_reqs;
280 struct completion wc_io_complete;
281 int wc_error;
282 };
283
284 #define O2HB_NEGO_TIMEOUT_MS (O2HB_MAX_WRITE_TIMEOUT_MS/2)
285
286 enum {
287 O2HB_NEGO_TIMEOUT_MSG = 1,
288 O2HB_NEGO_APPROVE_MSG = 2,
289 };
290
291 struct o2hb_nego_msg {
292 u8 node_num;
293 };
294
295 static void o2hb_write_timeout(struct work_struct *work)
296 {
297 int failed, quorum;
298 struct o2hb_region *reg =
299 container_of(work, struct o2hb_region,
300 hr_write_timeout_work.work);
301
302 mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
303 "milliseconds\n", reg->hr_dev_name,
304 jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
305
306 if (o2hb_global_heartbeat_active()) {
307 spin_lock(&o2hb_live_lock);
308 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
309 set_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
310 failed = bitmap_weight(o2hb_failed_region_bitmap,
311 O2NM_MAX_REGIONS);
312 quorum = bitmap_weight(o2hb_quorum_region_bitmap,
313 O2NM_MAX_REGIONS);
314 spin_unlock(&o2hb_live_lock);
315
316 mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n",
317 quorum, failed);
318
319 /*
320 * Fence if the number of failed regions >= half the number
321 * of quorum regions
322 */
323 if ((failed << 1) < quorum)
324 return;
325 }
326
327 o2quo_disk_timeout();
328 }
329
330 static void o2hb_arm_timeout(struct o2hb_region *reg)
331 {
332 /* Arm writeout only after thread reaches steady state */
333 if (atomic_read(®->hr_steady_iterations) != 0)
334 return;
335
336 mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n",
337 O2HB_MAX_WRITE_TIMEOUT_MS);
338
339 if (o2hb_global_heartbeat_active()) {
340 spin_lock(&o2hb_live_lock);
341 clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
342 spin_unlock(&o2hb_live_lock);
343 }
344 cancel_delayed_work(®->hr_write_timeout_work);
345 schedule_delayed_work(®->hr_write_timeout_work,
346 msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
347
348 cancel_delayed_work(®->hr_nego_timeout_work);
349 /* negotiate timeout must be less than write timeout. */
350 schedule_delayed_work(®->hr_nego_timeout_work,
351 msecs_to_jiffies(O2HB_NEGO_TIMEOUT_MS));
352 memset(reg->hr_nego_node_bitmap, 0, sizeof(reg->hr_nego_node_bitmap));
353 }
354
355 static void o2hb_disarm_timeout(struct o2hb_region *reg)
356 {
357 cancel_delayed_work_sync(®->hr_write_timeout_work);
358 cancel_delayed_work_sync(®->hr_nego_timeout_work);
359 }
360
361 static int o2hb_send_nego_msg(int key, int type, u8 target)
362 {
363 struct o2hb_nego_msg msg;
364 int status, ret;
365
366 msg.node_num = o2nm_this_node();
367 again:
368 ret = o2net_send_message(type, key, &msg, sizeof(msg),
369 target, &status);
370
371 if (ret == -EAGAIN || ret == -ENOMEM) {
372 msleep(100);
373 goto again;
374 }
375
376 return ret;
377 }
378
379 static void o2hb_nego_timeout(struct work_struct *work)
380 {
381 unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
382 int master_node, i, ret;
383 struct o2hb_region *reg;
384
385 reg = container_of(work, struct o2hb_region, hr_nego_timeout_work.work);
386 /* don't negotiate timeout if last hb failed since it is very
387 * possible io failed. Should let write timeout fence self.
388 */
389 if (reg->hr_last_hb_status)
390 return;
391
392 o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
393 /* lowest node as master node to make negotiate decision. */
394 master_node = find_next_bit(live_node_bitmap, O2NM_MAX_NODES, 0);
395
396 if (master_node == o2nm_this_node()) {
397 if (!test_bit(master_node, reg->hr_nego_node_bitmap)) {
398 printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s).\n",
399 o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000,
400 config_item_name(®->hr_item), reg->hr_dev_name);
401 set_bit(master_node, reg->hr_nego_node_bitmap);
402 }
403 if (memcmp(reg->hr_nego_node_bitmap, live_node_bitmap,
404 sizeof(reg->hr_nego_node_bitmap))) {
405 /* check negotiate bitmap every second to do timeout
406 * approve decision.
407 */
408 schedule_delayed_work(®->hr_nego_timeout_work,
409 msecs_to_jiffies(1000));
410
411 return;
412 }
413
414 printk(KERN_NOTICE "o2hb: all nodes hb write hung, maybe region %s (%s) is down.\n",
415 config_item_name(®->hr_item), reg->hr_dev_name);
416 /* approve negotiate timeout request. */
417 o2hb_arm_timeout(reg);
418
419 i = -1;
420 while ((i = find_next_bit(live_node_bitmap,
421 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
422 if (i == master_node)
423 continue;
424
425 mlog(ML_HEARTBEAT, "send NEGO_APPROVE msg to node %d\n", i);
426 ret = o2hb_send_nego_msg(reg->hr_key,
427 O2HB_NEGO_APPROVE_MSG, i);
428 if (ret)
429 mlog(ML_ERROR, "send NEGO_APPROVE msg to node %d fail %d\n",
430 i, ret);
431 }
432 } else {
433 /* negotiate timeout with master node. */
434 printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s), negotiate timeout with node %d.\n",
435 o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000, config_item_name(®->hr_item),
436 reg->hr_dev_name, master_node);
437 ret = o2hb_send_nego_msg(reg->hr_key, O2HB_NEGO_TIMEOUT_MSG,
438 master_node);
439 if (ret)
440 mlog(ML_ERROR, "send NEGO_TIMEOUT msg to node %d fail %d\n",
441 master_node, ret);
442 }
443 }
444
445 static int o2hb_nego_timeout_handler(struct o2net_msg *msg, u32 len, void *data,
446 void **ret_data)
447 {
448 struct o2hb_region *reg = data;
449 struct o2hb_nego_msg *nego_msg;
450
451 nego_msg = (struct o2hb_nego_msg *)msg->buf;
452 printk(KERN_NOTICE "o2hb: receive negotiate timeout message from node %d on region %s (%s).\n",
453 nego_msg->node_num, config_item_name(®->hr_item), reg->hr_dev_name);
454 if (nego_msg->node_num < O2NM_MAX_NODES)
455 set_bit(nego_msg->node_num, reg->hr_nego_node_bitmap);
456 else
457 mlog(ML_ERROR, "got nego timeout message from bad node.\n");
458
459 return 0;
460 }
461
462 static int o2hb_nego_approve_handler(struct o2net_msg *msg, u32 len, void *data,
463 void **ret_data)
464 {
465 struct o2hb_region *reg = data;
466
467 printk(KERN_NOTICE "o2hb: negotiate timeout approved by master node on region %s (%s).\n",
468 config_item_name(®->hr_item), reg->hr_dev_name);
469 o2hb_arm_timeout(reg);
470 return 0;
471 }
472
473 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc)
474 {
475 atomic_set(&wc->wc_num_reqs, 1);
476 init_completion(&wc->wc_io_complete);
477 wc->wc_error = 0;
478 }
479
480 /* Used in error paths too */
481 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
482 unsigned int num)
483 {
484 /* sadly atomic_sub_and_test() isn't available on all platforms. The
485 * good news is that the fast path only completes one at a time */
486 while(num--) {
487 if (atomic_dec_and_test(&wc->wc_num_reqs)) {
488 BUG_ON(num > 0);
489 complete(&wc->wc_io_complete);
490 }
491 }
492 }
493
494 static void o2hb_wait_on_io(struct o2hb_bio_wait_ctxt *wc)
495 {
496 o2hb_bio_wait_dec(wc, 1);
497 wait_for_completion(&wc->wc_io_complete);
498 }
499
500 static void o2hb_bio_end_io(struct bio *bio)
501 {
502 struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
503
504 if (bio->bi_status) {
505 mlog(ML_ERROR, "IO Error %d\n", bio->bi_status);
506 wc->wc_error = blk_status_to_errno(bio->bi_status);
507 }
508
509 o2hb_bio_wait_dec(wc, 1);
510 bio_put(bio);
511 }
512
513 /* Setup a Bio to cover I/O against num_slots slots starting at
514 * start_slot. */
515 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
516 struct o2hb_bio_wait_ctxt *wc,
517 unsigned int *current_slot,
518 unsigned int max_slots, int op,
519 int op_flags)
520 {
521 int len, current_page;
522 unsigned int vec_len, vec_start;
523 unsigned int bits = reg->hr_block_bits;
524 unsigned int spp = reg->hr_slots_per_page;
525 unsigned int cs = *current_slot;
526 struct bio *bio;
527 struct page *page;
528
529 /* Testing has shown this allocation to take long enough under
530 * GFP_KERNEL that the local node can get fenced. It would be
531 * nicest if we could pre-allocate these bios and avoid this
532 * all together. */
533 bio = bio_alloc(GFP_ATOMIC, 16);
534 if (!bio) {
535 mlog(ML_ERROR, "Could not alloc slots BIO!\n");
536 bio = ERR_PTR(-ENOMEM);
537 goto bail;
538 }
539
540 /* Must put everything in 512 byte sectors for the bio... */
541 bio->bi_iter.bi_sector = (reg->hr_start_block + cs) << (bits - 9);
542 bio_set_dev(bio, reg->hr_bdev);
543 bio->bi_private = wc;
544 bio->bi_end_io = o2hb_bio_end_io;
545 bio_set_op_attrs(bio, op, op_flags);
546
547 vec_start = (cs << bits) % PAGE_SIZE;
548 while(cs < max_slots) {
549 current_page = cs / spp;
550 page = reg->hr_slot_data[current_page];
551
552 vec_len = min(PAGE_SIZE - vec_start,
553 (max_slots-cs) * (PAGE_SIZE/spp) );
554
555 mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
556 current_page, vec_len, vec_start);
557
558 len = bio_add_page(bio, page, vec_len, vec_start);
559 if (len != vec_len) break;
560
561 cs += vec_len / (PAGE_SIZE/spp);
562 vec_start = 0;
563 }
564
565 bail:
566 *current_slot = cs;
567 return bio;
568 }
569
570 static int o2hb_read_slots(struct o2hb_region *reg,
571 unsigned int begin_slot,
572 unsigned int max_slots)
573 {
574 unsigned int current_slot = begin_slot;
575 int status;
576 struct o2hb_bio_wait_ctxt wc;
577 struct bio *bio;
578
579 o2hb_bio_wait_init(&wc);
580
581 while(current_slot < max_slots) {
582 bio = o2hb_setup_one_bio(reg, &wc, ¤t_slot, max_slots,
583 REQ_OP_READ, 0);
584 if (IS_ERR(bio)) {
585 status = PTR_ERR(bio);
586 mlog_errno(status);
587 goto bail_and_wait;
588 }
589
590 atomic_inc(&wc.wc_num_reqs);
591 submit_bio(bio);
592 }
593
594 status = 0;
595
596 bail_and_wait:
597 o2hb_wait_on_io(&wc);
598 if (wc.wc_error && !status)
599 status = wc.wc_error;
600
601 return status;
602 }
603
604 static int o2hb_issue_node_write(struct o2hb_region *reg,
605 struct o2hb_bio_wait_ctxt *write_wc)
606 {
607 int status;
608 unsigned int slot;
609 struct bio *bio;
610
611 o2hb_bio_wait_init(write_wc);
612
613 slot = o2nm_this_node();
614
615 bio = o2hb_setup_one_bio(reg, write_wc, &slot, slot+1, REQ_OP_WRITE,
616 REQ_SYNC);
617 if (IS_ERR(bio)) {
618 status = PTR_ERR(bio);
619 mlog_errno(status);
620 goto bail;
621 }
622
623 atomic_inc(&write_wc->wc_num_reqs);
624 submit_bio(bio);
625
626 status = 0;
627 bail:
628 return status;
629 }
630
631 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
632 struct o2hb_disk_heartbeat_block *hb_block)
633 {
634 __le32 old_cksum;
635 u32 ret;
636
637 /* We want to compute the block crc with a 0 value in the
638 * hb_cksum field. Save it off here and replace after the
639 * crc. */
640 old_cksum = hb_block->hb_cksum;
641 hb_block->hb_cksum = 0;
642
643 ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
644
645 hb_block->hb_cksum = old_cksum;
646
647 return ret;
648 }
649
650 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
651 {
652 mlog(ML_ERROR, "Dump slot information: seq = 0x%llx, node = %u, "
653 "cksum = 0x%x, generation 0x%llx\n",
654 (long long)le64_to_cpu(hb_block->hb_seq),
655 hb_block->hb_node, le32_to_cpu(hb_block->hb_cksum),
656 (long long)le64_to_cpu(hb_block->hb_generation));
657 }
658
659 static int o2hb_verify_crc(struct o2hb_region *reg,
660 struct o2hb_disk_heartbeat_block *hb_block)
661 {
662 u32 read, computed;
663
664 read = le32_to_cpu(hb_block->hb_cksum);
665 computed = o2hb_compute_block_crc_le(reg, hb_block);
666
667 return read == computed;
668 }
669
670 /*
671 * Compare the slot data with what we wrote in the last iteration.
672 * If the match fails, print an appropriate error message. This is to
673 * detect errors like... another node hearting on the same slot,
674 * flaky device that is losing writes, etc.
675 * Returns 1 if check succeeds, 0 otherwise.
676 */
677 static int o2hb_check_own_slot(struct o2hb_region *reg)
678 {
679 struct o2hb_disk_slot *slot;
680 struct o2hb_disk_heartbeat_block *hb_block;
681 char *errstr;
682
683 slot = ®->hr_slots[o2nm_this_node()];
684 /* Don't check on our 1st timestamp */
685 if (!slot->ds_last_time)
686 return 0;
687
688 hb_block = slot->ds_raw_block;
689 if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time &&
690 le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation &&
691 hb_block->hb_node == slot->ds_node_num)
692 return 1;
693
694 #define ERRSTR1 "Another node is heartbeating on device"
695 #define ERRSTR2 "Heartbeat generation mismatch on device"
696 #define ERRSTR3 "Heartbeat sequence mismatch on device"
697
698 if (hb_block->hb_node != slot->ds_node_num)
699 errstr = ERRSTR1;
700 else if (le64_to_cpu(hb_block->hb_generation) !=
701 slot->ds_last_generation)
702 errstr = ERRSTR2;
703 else
704 errstr = ERRSTR3;
705
706 mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), "
707 "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name,
708 slot->ds_node_num, (unsigned long long)slot->ds_last_generation,
709 (unsigned long long)slot->ds_last_time, hb_block->hb_node,
710 (unsigned long long)le64_to_cpu(hb_block->hb_generation),
711 (unsigned long long)le64_to_cpu(hb_block->hb_seq));
712
713 return 0;
714 }
715
716 static inline void o2hb_prepare_block(struct o2hb_region *reg,
717 u64 generation)
718 {
719 int node_num;
720 u64 cputime;
721 struct o2hb_disk_slot *slot;
722 struct o2hb_disk_heartbeat_block *hb_block;
723
724 node_num = o2nm_this_node();
725 slot = ®->hr_slots[node_num];
726
727 hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
728 memset(hb_block, 0, reg->hr_block_bytes);
729 /* TODO: time stuff */
730 cputime = ktime_get_real_seconds();
731 if (!cputime)
732 cputime = 1;
733
734 hb_block->hb_seq = cpu_to_le64(cputime);
735 hb_block->hb_node = node_num;
736 hb_block->hb_generation = cpu_to_le64(generation);
737 hb_block->hb_dead_ms = cpu_to_le32(o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS);
738
739 /* This step must always happen last! */
740 hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
741 hb_block));
742
743 mlog(ML_HB_BIO, "our node generation = 0x%llx, cksum = 0x%x\n",
744 (long long)generation,
745 le32_to_cpu(hb_block->hb_cksum));
746 }
747
748 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
749 struct o2nm_node *node,
750 int idx)
751 {
752 struct o2hb_callback_func *f;
753
754 list_for_each_entry(f, &hbcall->list, hc_item) {
755 mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
756 (f->hc_func)(node, idx, f->hc_data);
757 }
758 }
759
760 /* Will run the list in order until we process the passed event */
761 static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
762 {
763 struct o2hb_callback *hbcall;
764 struct o2hb_node_event *event;
765
766 /* Holding callback sem assures we don't alter the callback
767 * lists when doing this, and serializes ourselves with other
768 * processes wanting callbacks. */
769 down_write(&o2hb_callback_sem);
770
771 spin_lock(&o2hb_live_lock);
772 while (!list_empty(&o2hb_node_events)
773 && !list_empty(&queued_event->hn_item)) {
774 event = list_entry(o2hb_node_events.next,
775 struct o2hb_node_event,
776 hn_item);
777 list_del_init(&event->hn_item);
778 spin_unlock(&o2hb_live_lock);
779
780 mlog(ML_HEARTBEAT, "Node %s event for %d\n",
781 event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
782 event->hn_node_num);
783
784 hbcall = hbcall_from_type(event->hn_event_type);
785
786 /* We should *never* have gotten on to the list with a
787 * bad type... This isn't something that we should try
788 * to recover from. */
789 BUG_ON(IS_ERR(hbcall));
790
791 o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
792
793 spin_lock(&o2hb_live_lock);
794 }
795 spin_unlock(&o2hb_live_lock);
796
797 up_write(&o2hb_callback_sem);
798 }
799
800 static void o2hb_queue_node_event(struct o2hb_node_event *event,
801 enum o2hb_callback_type type,
802 struct o2nm_node *node,
803 int node_num)
804 {
805 assert_spin_locked(&o2hb_live_lock);
806
807 BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB));
808
809 event->hn_event_type = type;
810 event->hn_node = node;
811 event->hn_node_num = node_num;
812
813 mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
814 type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
815
816 list_add_tail(&event->hn_item, &o2hb_node_events);
817 }
818
819 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
820 {
821 struct o2hb_node_event event =
822 { .hn_item = LIST_HEAD_INIT(event.hn_item), };
823 struct o2nm_node *node;
824 int queued = 0;
825
826 node = o2nm_get_node_by_num(slot->ds_node_num);
827 if (!node)
828 return;
829
830 spin_lock(&o2hb_live_lock);
831 if (!list_empty(&slot->ds_live_item)) {
832 mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
833 slot->ds_node_num);
834
835 list_del_init(&slot->ds_live_item);
836
837 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
838 clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
839
840 o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
841 slot->ds_node_num);
842 queued = 1;
843 }
844 }
845 spin_unlock(&o2hb_live_lock);
846
847 if (queued)
848 o2hb_run_event_list(&event);
849
850 o2nm_node_put(node);
851 }
852
853 static void o2hb_set_quorum_device(struct o2hb_region *reg)
854 {
855 if (!o2hb_global_heartbeat_active())
856 return;
857
858 /* Prevent race with o2hb_heartbeat_group_drop_item() */
859 if (kthread_should_stop())
860 return;
861
862 /* Tag region as quorum only after thread reaches steady state */
863 if (atomic_read(®->hr_steady_iterations) != 0)
864 return;
865
866 spin_lock(&o2hb_live_lock);
867
868 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
869 goto unlock;
870
871 /*
872 * A region can be added to the quorum only when it sees all
873 * live nodes heartbeat on it. In other words, the region has been
874 * added to all nodes.
875 */
876 if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap,
877 sizeof(o2hb_live_node_bitmap)))
878 goto unlock;
879
880 printk(KERN_NOTICE "o2hb: Region %s (%s) is now a quorum device\n",
881 config_item_name(®->hr_item), reg->hr_dev_name);
882
883 set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
884
885 /*
886 * If global heartbeat active, unpin all regions if the
887 * region count > CUT_OFF
888 */
889 if (bitmap_weight(o2hb_quorum_region_bitmap,
890 O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF)
891 o2hb_region_unpin(NULL);
892 unlock:
893 spin_unlock(&o2hb_live_lock);
894 }
895
896 static int o2hb_check_slot(struct o2hb_region *reg,
897 struct o2hb_disk_slot *slot)
898 {
899 int changed = 0, gen_changed = 0;
900 struct o2hb_node_event event =
901 { .hn_item = LIST_HEAD_INIT(event.hn_item), };
902 struct o2nm_node *node;
903 struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
904 u64 cputime;
905 unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS;
906 unsigned int slot_dead_ms;
907 int tmp;
908 int queued = 0;
909
910 memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
911
912 /*
913 * If a node is no longer configured but is still in the livemap, we
914 * may need to clear that bit from the livemap.
915 */
916 node = o2nm_get_node_by_num(slot->ds_node_num);
917 if (!node) {
918 spin_lock(&o2hb_live_lock);
919 tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap);
920 spin_unlock(&o2hb_live_lock);
921 if (!tmp)
922 return 0;
923 }
924
925 if (!o2hb_verify_crc(reg, hb_block)) {
926 /* all paths from here will drop o2hb_live_lock for
927 * us. */
928 spin_lock(&o2hb_live_lock);
929
930 /* Don't print an error on the console in this case -
931 * a freshly formatted heartbeat area will not have a
932 * crc set on it. */
933 if (list_empty(&slot->ds_live_item))
934 goto out;
935
936 /* The node is live but pushed out a bad crc. We
937 * consider it a transient miss but don't populate any
938 * other values as they may be junk. */
939 mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
940 slot->ds_node_num, reg->hr_dev_name);
941 o2hb_dump_slot(hb_block);
942
943 slot->ds_equal_samples++;
944 goto fire_callbacks;
945 }
946
947 /* we don't care if these wrap.. the state transitions below
948 * clear at the right places */
949 cputime = le64_to_cpu(hb_block->hb_seq);
950 if (slot->ds_last_time != cputime)
951 slot->ds_changed_samples++;
952 else
953 slot->ds_equal_samples++;
954 slot->ds_last_time = cputime;
955
956 /* The node changed heartbeat generations. We assume this to
957 * mean it dropped off but came back before we timed out. We
958 * want to consider it down for the time being but don't want
959 * to lose any changed_samples state we might build up to
960 * considering it live again. */
961 if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
962 gen_changed = 1;
963 slot->ds_equal_samples = 0;
964 mlog(ML_HEARTBEAT, "Node %d changed generation (0x%llx "
965 "to 0x%llx)\n", slot->ds_node_num,
966 (long long)slot->ds_last_generation,
967 (long long)le64_to_cpu(hb_block->hb_generation));
968 }
969
970 slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
971
972 mlog(ML_HEARTBEAT, "Slot %d gen 0x%llx cksum 0x%x "
973 "seq %llu last %llu changed %u equal %u\n",
974 slot->ds_node_num, (long long)slot->ds_last_generation,
975 le32_to_cpu(hb_block->hb_cksum),
976 (unsigned long long)le64_to_cpu(hb_block->hb_seq),
977 (unsigned long long)slot->ds_last_time, slot->ds_changed_samples,
978 slot->ds_equal_samples);
979
980 spin_lock(&o2hb_live_lock);
981
982 fire_callbacks:
983 /* dead nodes only come to life after some number of
984 * changes at any time during their dead time */
985 if (list_empty(&slot->ds_live_item) &&
986 slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
987 mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n",
988 slot->ds_node_num, (long long)slot->ds_last_generation);
989
990 set_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
991
992 /* first on the list generates a callback */
993 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
994 mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes "
995 "bitmap\n", slot->ds_node_num);
996 set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
997
998 o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
999 slot->ds_node_num);
1000
1001 changed = 1;
1002 queued = 1;
1003 }
1004
1005 list_add_tail(&slot->ds_live_item,
1006 &o2hb_live_slots[slot->ds_node_num]);
1007
1008 slot->ds_equal_samples = 0;
1009
1010 /* We want to be sure that all nodes agree on the
1011 * number of milliseconds before a node will be
1012 * considered dead. The self-fencing timeout is
1013 * computed from this value, and a discrepancy might
1014 * result in heartbeat calling a node dead when it
1015 * hasn't self-fenced yet. */
1016 slot_dead_ms = le32_to_cpu(hb_block->hb_dead_ms);
1017 if (slot_dead_ms && slot_dead_ms != dead_ms) {
1018 /* TODO: Perhaps we can fail the region here. */
1019 mlog(ML_ERROR, "Node %d on device %s has a dead count "
1020 "of %u ms, but our count is %u ms.\n"
1021 "Please double check your configuration values "
1022 "for 'O2CB_HEARTBEAT_THRESHOLD'\n",
1023 slot->ds_node_num, reg->hr_dev_name, slot_dead_ms,
1024 dead_ms);
1025 }
1026 goto out;
1027 }
1028
1029 /* if the list is dead, we're done.. */
1030 if (list_empty(&slot->ds_live_item))
1031 goto out;
1032
1033 /* live nodes only go dead after enough consequtive missed
1034 * samples.. reset the missed counter whenever we see
1035 * activity */
1036 if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
1037 mlog(ML_HEARTBEAT, "Node %d left my region\n",
1038 slot->ds_node_num);
1039
1040 clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
1041
1042 /* last off the live_slot generates a callback */
1043 list_del_init(&slot->ds_live_item);
1044 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
1045 mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live "
1046 "nodes bitmap\n", slot->ds_node_num);
1047 clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
1048
1049 /* node can be null */
1050 o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB,
1051 node, slot->ds_node_num);
1052
1053 changed = 1;
1054 queued = 1;
1055 }
1056
1057 /* We don't clear this because the node is still
1058 * actually writing new blocks. */
1059 if (!gen_changed)
1060 slot->ds_changed_samples = 0;
1061 goto out;
1062 }
1063 if (slot->ds_changed_samples) {
1064 slot->ds_changed_samples = 0;
1065 slot->ds_equal_samples = 0;
1066 }
1067 out:
1068 spin_unlock(&o2hb_live_lock);
1069
1070 if (queued)
1071 o2hb_run_event_list(&event);
1072
1073 if (node)
1074 o2nm_node_put(node);
1075 return changed;
1076 }
1077
1078 static int o2hb_highest_node(unsigned long *nodes, int numbits)
1079 {
1080 return find_last_bit(nodes, numbits);
1081 }
1082
1083 static int o2hb_lowest_node(unsigned long *nodes, int numbits)
1084 {
1085 return find_first_bit(nodes, numbits);
1086 }
1087
1088 static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
1089 {
1090 int i, ret, highest_node, lowest_node;
1091 int membership_change = 0, own_slot_ok = 0;
1092 unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
1093 unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
1094 struct o2hb_bio_wait_ctxt write_wc;
1095
1096 ret = o2nm_configured_node_map(configured_nodes,
1097 sizeof(configured_nodes));
1098 if (ret) {
1099 mlog_errno(ret);
1100 goto bail;
1101 }
1102
1103 /*
1104 * If a node is not configured but is in the livemap, we still need
1105 * to read the slot so as to be able to remove it from the livemap.
1106 */
1107 o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
1108 i = -1;
1109 while ((i = find_next_bit(live_node_bitmap,
1110 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1111 set_bit(i, configured_nodes);
1112 }
1113
1114 highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
1115 lowest_node = o2hb_lowest_node(configured_nodes, O2NM_MAX_NODES);
1116 if (highest_node >= O2NM_MAX_NODES || lowest_node >= O2NM_MAX_NODES) {
1117 mlog(ML_NOTICE, "o2hb: No configured nodes found!\n");
1118 ret = -EINVAL;
1119 goto bail;
1120 }
1121
1122 /* No sense in reading the slots of nodes that don't exist
1123 * yet. Of course, if the node definitions have holes in them
1124 * then we're reading an empty slot anyway... Consider this
1125 * best-effort. */
1126 ret = o2hb_read_slots(reg, lowest_node, highest_node + 1);
1127 if (ret < 0) {
1128 mlog_errno(ret);
1129 goto bail;
1130 }
1131
1132 /* With an up to date view of the slots, we can check that no
1133 * other node has been improperly configured to heartbeat in
1134 * our slot. */
1135 own_slot_ok = o2hb_check_own_slot(reg);
1136
1137 /* fill in the proper info for our next heartbeat */
1138 o2hb_prepare_block(reg, reg->hr_generation);
1139
1140 ret = o2hb_issue_node_write(reg, &write_wc);
1141 if (ret < 0) {
1142 mlog_errno(ret);
1143 goto bail;
1144 }
1145
1146 i = -1;
1147 while((i = find_next_bit(configured_nodes,
1148 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1149 membership_change |= o2hb_check_slot(reg, ®->hr_slots[i]);
1150 }
1151
1152 /*
1153 * We have to be sure we've advertised ourselves on disk
1154 * before we can go to steady state. This ensures that
1155 * people we find in our steady state have seen us.
1156 */
1157 o2hb_wait_on_io(&write_wc);
1158 if (write_wc.wc_error) {
1159 /* Do not re-arm the write timeout on I/O error - we
1160 * can't be sure that the new block ever made it to
1161 * disk */
1162 mlog(ML_ERROR, "Write error %d on device \"%s\"\n",
1163 write_wc.wc_error, reg->hr_dev_name);
1164 ret = write_wc.wc_error;
1165 goto bail;
1166 }
1167
1168 /* Skip disarming the timeout if own slot has stale/bad data */
1169 if (own_slot_ok) {
1170 o2hb_set_quorum_device(reg);
1171 o2hb_arm_timeout(reg);
1172 reg->hr_last_timeout_start = jiffies;
1173 }
1174
1175 bail:
1176 /* let the person who launched us know when things are steady */
1177 if (atomic_read(®->hr_steady_iterations) != 0) {
1178 if (!ret && own_slot_ok && !membership_change) {
1179 if (atomic_dec_and_test(®->hr_steady_iterations))
1180 wake_up(&o2hb_steady_queue);
1181 }
1182 }
1183
1184 if (atomic_read(®->hr_steady_iterations) != 0) {
1185 if (atomic_dec_and_test(®->hr_unsteady_iterations)) {
1186 printk(KERN_NOTICE "o2hb: Unable to stabilize "
1187 "heartbeart on region %s (%s)\n",
1188 config_item_name(®->hr_item),
1189 reg->hr_dev_name);
1190 atomic_set(®->hr_steady_iterations, 0);
1191 reg->hr_aborted_start = 1;
1192 wake_up(&o2hb_steady_queue);
1193 ret = -EIO;
1194 }
1195 }
1196
1197 return ret;
1198 }
1199
1200 /*
1201 * we ride the region ref that the region dir holds. before the region
1202 * dir is removed and drops it ref it will wait to tear down this
1203 * thread.
1204 */
1205 static int o2hb_thread(void *data)
1206 {
1207 int i, ret;
1208 struct o2hb_region *reg = data;
1209 struct o2hb_bio_wait_ctxt write_wc;
1210 ktime_t before_hb, after_hb;
1211 unsigned int elapsed_msec;
1212
1213 mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
1214
1215 set_user_nice(current, MIN_NICE);
1216
1217 /* Pin node */
1218 ret = o2nm_depend_this_node();
1219 if (ret) {
1220 mlog(ML_ERROR, "Node has been deleted, ret = %d\n", ret);
1221 reg->hr_node_deleted = 1;
1222 wake_up(&o2hb_steady_queue);
1223 return 0;
1224 }
1225
1226 while (!kthread_should_stop() &&
1227 !reg->hr_unclean_stop && !reg->hr_aborted_start) {
1228 /* We track the time spent inside
1229 * o2hb_do_disk_heartbeat so that we avoid more than
1230 * hr_timeout_ms between disk writes. On busy systems
1231 * this should result in a heartbeat which is less
1232 * likely to time itself out. */
1233 before_hb = ktime_get_real();
1234
1235 ret = o2hb_do_disk_heartbeat(reg);
1236 reg->hr_last_hb_status = ret;
1237
1238 after_hb = ktime_get_real();
1239
1240 elapsed_msec = (unsigned int)
1241 ktime_ms_delta(after_hb, before_hb);
1242
1243 mlog(ML_HEARTBEAT,
1244 "start = %lld, end = %lld, msec = %u, ret = %d\n",
1245 before_hb, after_hb, elapsed_msec, ret);
1246
1247 if (!kthread_should_stop() &&
1248 elapsed_msec < reg->hr_timeout_ms) {
1249 /* the kthread api has blocked signals for us so no
1250 * need to record the return value. */
1251 msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
1252 }
1253 }
1254
1255 o2hb_disarm_timeout(reg);
1256
1257 /* unclean stop is only used in very bad situation */
1258 for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
1259 o2hb_shutdown_slot(®->hr_slots[i]);
1260
1261 /* Explicit down notification - avoid forcing the other nodes
1262 * to timeout on this region when we could just as easily
1263 * write a clear generation - thus indicating to them that
1264 * this node has left this region.
1265 */
1266 if (!reg->hr_unclean_stop && !reg->hr_aborted_start) {
1267 o2hb_prepare_block(reg, 0);
1268 ret = o2hb_issue_node_write(reg, &write_wc);
1269 if (ret == 0)
1270 o2hb_wait_on_io(&write_wc);
1271 else
1272 mlog_errno(ret);
1273 }
1274
1275 /* Unpin node */
1276 o2nm_undepend_this_node();
1277
1278 mlog(ML_HEARTBEAT|ML_KTHREAD, "o2hb thread exiting\n");
1279
1280 return 0;
1281 }
1282
1283 #ifdef CONFIG_DEBUG_FS
1284 static int o2hb_debug_open(struct inode *inode, struct file *file)
1285 {
1286 struct o2hb_debug_buf *db = inode->i_private;
1287 struct o2hb_region *reg;
1288 unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1289 unsigned long lts;
1290 char *buf = NULL;
1291 int i = -1;
1292 int out = 0;
1293
1294 /* max_nodes should be the largest bitmap we pass here */
1295 BUG_ON(sizeof(map) < db->db_size);
1296
1297 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
1298 if (!buf)
1299 goto bail;
1300
1301 switch (db->db_type) {
1302 case O2HB_DB_TYPE_LIVENODES:
1303 case O2HB_DB_TYPE_LIVEREGIONS:
1304 case O2HB_DB_TYPE_QUORUMREGIONS:
1305 case O2HB_DB_TYPE_FAILEDREGIONS:
1306 spin_lock(&o2hb_live_lock);
1307 memcpy(map, db->db_data, db->db_size);
1308 spin_unlock(&o2hb_live_lock);
1309 break;
1310
1311 case O2HB_DB_TYPE_REGION_LIVENODES:
1312 spin_lock(&o2hb_live_lock);
1313 reg = (struct o2hb_region *)db->db_data;
1314 memcpy(map, reg->hr_live_node_bitmap, db->db_size);
1315 spin_unlock(&o2hb_live_lock);
1316 break;
1317
1318 case O2HB_DB_TYPE_REGION_NUMBER:
1319 reg = (struct o2hb_region *)db->db_data;
1320 out += snprintf(buf + out, PAGE_SIZE - out, "%d\n",
1321 reg->hr_region_num);
1322 goto done;
1323
1324 case O2HB_DB_TYPE_REGION_ELAPSED_TIME:
1325 reg = (struct o2hb_region *)db->db_data;
1326 lts = reg->hr_last_timeout_start;
1327 /* If 0, it has never been set before */
1328 if (lts)
1329 lts = jiffies_to_msecs(jiffies - lts);
1330 out += snprintf(buf + out, PAGE_SIZE - out, "%lu\n", lts);
1331 goto done;
1332
1333 case O2HB_DB_TYPE_REGION_PINNED:
1334 reg = (struct o2hb_region *)db->db_data;
1335 out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
1336 !!reg->hr_item_pinned);
1337 goto done;
1338
1339 default:
1340 goto done;
1341 }
1342
1343 while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len)
1344 out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i);
1345 out += snprintf(buf + out, PAGE_SIZE - out, "\n");
1346
1347 done:
1348 i_size_write(inode, out);
1349
1350 file->private_data = buf;
1351
1352 return 0;
1353 bail:
1354 return -ENOMEM;
1355 }
1356
1357 static int o2hb_debug_release(struct inode *inode, struct file *file)
1358 {
1359 kfree(file->private_data);
1360 return 0;
1361 }
1362
1363 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1364 size_t nbytes, loff_t *ppos)
1365 {
1366 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
1367 i_size_read(file->f_mapping->host));
1368 }
1369 #else
1370 static int o2hb_debug_open(struct inode *inode, struct file *file)
1371 {
1372 return 0;
1373 }
1374 static int o2hb_debug_release(struct inode *inode, struct file *file)
1375 {
1376 return 0;
1377 }
1378 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1379 size_t nbytes, loff_t *ppos)
1380 {
1381 return 0;
1382 }
1383 #endif /* CONFIG_DEBUG_FS */
1384
1385 static const struct file_operations o2hb_debug_fops = {
1386 .open = o2hb_debug_open,
1387 .release = o2hb_debug_release,
1388 .read = o2hb_debug_read,
1389 .llseek = generic_file_llseek,
1390 };
1391
1392 void o2hb_exit(void)
1393 {
1394 debugfs_remove(o2hb_debug_failedregions);
1395 debugfs_remove(o2hb_debug_quorumregions);
1396 debugfs_remove(o2hb_debug_liveregions);
1397 debugfs_remove(o2hb_debug_livenodes);
1398 debugfs_remove(o2hb_debug_dir);
1399 kfree(o2hb_db_livenodes);
1400 kfree(o2hb_db_liveregions);
1401 kfree(o2hb_db_quorumregions);
1402 kfree(o2hb_db_failedregions);
1403 }
1404
1405 static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir,
1406 struct o2hb_debug_buf **db, int db_len,
1407 int type, int size, int len, void *data)
1408 {
1409 *db = kmalloc(db_len, GFP_KERNEL);
1410 if (!*db)
1411 return NULL;
1412
1413 (*db)->db_type = type;
1414 (*db)->db_size = size;
1415 (*db)->db_len = len;
1416 (*db)->db_data = data;
1417
1418 return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db,
1419 &o2hb_debug_fops);
1420 }
1421
1422 static int o2hb_debug_init(void)
1423 {
1424 int ret = -ENOMEM;
1425
1426 o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL);
1427 if (!o2hb_debug_dir) {
1428 mlog_errno(ret);
1429 goto bail;
1430 }
1431
1432 o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES,
1433 o2hb_debug_dir,
1434 &o2hb_db_livenodes,
1435 sizeof(*o2hb_db_livenodes),
1436 O2HB_DB_TYPE_LIVENODES,
1437 sizeof(o2hb_live_node_bitmap),
1438 O2NM_MAX_NODES,
1439 o2hb_live_node_bitmap);
1440 if (!o2hb_debug_livenodes) {
1441 mlog_errno(ret);
1442 goto bail;
1443 }
1444
1445 o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS,
1446 o2hb_debug_dir,
1447 &o2hb_db_liveregions,
1448 sizeof(*o2hb_db_liveregions),
1449 O2HB_DB_TYPE_LIVEREGIONS,
1450 sizeof(o2hb_live_region_bitmap),
1451 O2NM_MAX_REGIONS,
1452 o2hb_live_region_bitmap);
1453 if (!o2hb_debug_liveregions) {
1454 mlog_errno(ret);
1455 goto bail;
1456 }
1457
1458 o2hb_debug_quorumregions =
1459 o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS,
1460 o2hb_debug_dir,
1461 &o2hb_db_quorumregions,
1462 sizeof(*o2hb_db_quorumregions),
1463 O2HB_DB_TYPE_QUORUMREGIONS,
1464 sizeof(o2hb_quorum_region_bitmap),
1465 O2NM_MAX_REGIONS,
1466 o2hb_quorum_region_bitmap);
1467 if (!o2hb_debug_quorumregions) {
1468 mlog_errno(ret);
1469 goto bail;
1470 }
1471
1472 o2hb_debug_failedregions =
1473 o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS,
1474 o2hb_debug_dir,
1475 &o2hb_db_failedregions,
1476 sizeof(*o2hb_db_failedregions),
1477 O2HB_DB_TYPE_FAILEDREGIONS,
1478 sizeof(o2hb_failed_region_bitmap),
1479 O2NM_MAX_REGIONS,
1480 o2hb_failed_region_bitmap);
1481 if (!o2hb_debug_failedregions) {
1482 mlog_errno(ret);
1483 goto bail;
1484 }
1485
1486 ret = 0;
1487 bail:
1488 if (ret)
1489 o2hb_exit();
1490
1491 return ret;
1492 }
1493
1494 int o2hb_init(void)
1495 {
1496 int i;
1497
1498 for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
1499 INIT_LIST_HEAD(&o2hb_callbacks[i].list);
1500
1501 for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
1502 INIT_LIST_HEAD(&o2hb_live_slots[i]);
1503
1504 INIT_LIST_HEAD(&o2hb_node_events);
1505
1506 memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
1507 memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap));
1508 memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap));
1509 memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap));
1510 memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap));
1511
1512 o2hb_dependent_users = 0;
1513
1514 return o2hb_debug_init();
1515 }
1516
1517 /* if we're already in a callback then we're already serialized by the sem */
1518 static void o2hb_fill_node_map_from_callback(unsigned long *map,
1519 unsigned bytes)
1520 {
1521 BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
1522
1523 memcpy(map, &o2hb_live_node_bitmap, bytes);
1524 }
1525
1526 /*
1527 * get a map of all nodes that are heartbeating in any regions
1528 */
1529 void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
1530 {
1531 /* callers want to serialize this map and callbacks so that they
1532 * can trust that they don't miss nodes coming to the party */
1533 down_read(&o2hb_callback_sem);
1534 spin_lock(&o2hb_live_lock);
1535 o2hb_fill_node_map_from_callback(map, bytes);
1536 spin_unlock(&o2hb_live_lock);
1537 up_read(&o2hb_callback_sem);
1538 }
1539 EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
1540
1541 /*
1542 * heartbeat configfs bits. The heartbeat set is a default set under
1543 * the cluster set in nodemanager.c.
1544 */
1545
1546 static struct o2hb_region *to_o2hb_region(struct config_item *item)
1547 {
1548 return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
1549 }
1550
1551 /* drop_item only drops its ref after killing the thread, nothing should
1552 * be using the region anymore. this has to clean up any state that
1553 * attributes might have built up. */
1554 static void o2hb_region_release(struct config_item *item)
1555 {
1556 int i;
1557 struct page *page;
1558 struct o2hb_region *reg = to_o2hb_region(item);
1559
1560 mlog(ML_HEARTBEAT, "hb region release (%s)\n", reg->hr_dev_name);
1561
1562 kfree(reg->hr_tmp_block);
1563
1564 if (reg->hr_slot_data) {
1565 for (i = 0; i < reg->hr_num_pages; i++) {
1566 page = reg->hr_slot_data[i];
1567 if (page)
1568 __free_page(page);
1569 }
1570 kfree(reg->hr_slot_data);
1571 }
1572
1573 if (reg->hr_bdev)
1574 blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1575
1576 kfree(reg->hr_slots);
1577
1578 debugfs_remove(reg->hr_debug_livenodes);
1579 debugfs_remove(reg->hr_debug_regnum);
1580 debugfs_remove(reg->hr_debug_elapsed_time);
1581 debugfs_remove(reg->hr_debug_pinned);
1582 debugfs_remove(reg->hr_debug_dir);
1583 kfree(reg->hr_db_livenodes);
1584 kfree(reg->hr_db_regnum);
1585 kfree(reg->hr_db_elapsed_time);
1586 kfree(reg->hr_db_pinned);
1587
1588 spin_lock(&o2hb_live_lock);
1589 list_del(®->hr_all_item);
1590 spin_unlock(&o2hb_live_lock);
1591
1592 o2net_unregister_handler_list(®->hr_handler_list);
1593 kfree(reg);
1594 }
1595
1596 static int o2hb_read_block_input(struct o2hb_region *reg,
1597 const char *page,
1598 unsigned long *ret_bytes,
1599 unsigned int *ret_bits)
1600 {
1601 unsigned long bytes;
1602 char *p = (char *)page;
1603
1604 bytes = simple_strtoul(p, &p, 0);
1605 if (!p || (*p && (*p != '\n')))
1606 return -EINVAL;
1607
1608 /* Heartbeat and fs min / max block sizes are the same. */
1609 if (bytes > 4096 || bytes < 512)
1610 return -ERANGE;
1611 if (hweight16(bytes) != 1)
1612 return -EINVAL;
1613
1614 if (ret_bytes)
1615 *ret_bytes = bytes;
1616 if (ret_bits)
1617 *ret_bits = ffs(bytes) - 1;
1618
1619 return 0;
1620 }
1621
1622 static ssize_t o2hb_region_block_bytes_show(struct config_item *item,
1623 char *page)
1624 {
1625 return sprintf(page, "%u\n", to_o2hb_region(item)->hr_block_bytes);
1626 }
1627
1628 static ssize_t o2hb_region_block_bytes_store(struct config_item *item,
1629 const char *page,
1630 size_t count)
1631 {
1632 struct o2hb_region *reg = to_o2hb_region(item);
1633 int status;
1634 unsigned long block_bytes;
1635 unsigned int block_bits;
1636
1637 if (reg->hr_bdev)
1638 return -EINVAL;
1639
1640 status = o2hb_read_block_input(reg, page, &block_bytes,
1641 &block_bits);
1642 if (status)
1643 return status;
1644
1645 reg->hr_block_bytes = (unsigned int)block_bytes;
1646 reg->hr_block_bits = block_bits;
1647
1648 return count;
1649 }
1650
1651 static ssize_t o2hb_region_start_block_show(struct config_item *item,
1652 char *page)
1653 {
1654 return sprintf(page, "%llu\n", to_o2hb_region(item)->hr_start_block);
1655 }
1656
1657 static ssize_t o2hb_region_start_block_store(struct config_item *item,
1658 const char *page,
1659 size_t count)
1660 {
1661 struct o2hb_region *reg = to_o2hb_region(item);
1662 unsigned long long tmp;
1663 char *p = (char *)page;
1664
1665 if (reg->hr_bdev)
1666 return -EINVAL;
1667
1668 tmp = simple_strtoull(p, &p, 0);
1669 if (!p || (*p && (*p != '\n')))
1670 return -EINVAL;
1671
1672 reg->hr_start_block = tmp;
1673
1674 return count;
1675 }
1676
1677 static ssize_t o2hb_region_blocks_show(struct config_item *item, char *page)
1678 {
1679 return sprintf(page, "%d\n", to_o2hb_region(item)->hr_blocks);
1680 }
1681
1682 static ssize_t o2hb_region_blocks_store(struct config_item *item,
1683 const char *page,
1684 size_t count)
1685 {
1686 struct o2hb_region *reg = to_o2hb_region(item);
1687 unsigned long tmp;
1688 char *p = (char *)page;
1689
1690 if (reg->hr_bdev)
1691 return -EINVAL;
1692
1693 tmp = simple_strtoul(p, &p, 0);
1694 if (!p || (*p && (*p != '\n')))
1695 return -EINVAL;
1696
1697 if (tmp > O2NM_MAX_NODES || tmp == 0)
1698 return -ERANGE;
1699
1700 reg->hr_blocks = (unsigned int)tmp;
1701
1702 return count;
1703 }
1704
1705 static ssize_t o2hb_region_dev_show(struct config_item *item, char *page)
1706 {
1707 unsigned int ret = 0;
1708
1709 if (to_o2hb_region(item)->hr_bdev)
1710 ret = sprintf(page, "%s\n", to_o2hb_region(item)->hr_dev_name);
1711
1712 return ret;
1713 }
1714
1715 static void o2hb_init_region_params(struct o2hb_region *reg)
1716 {
1717 reg->hr_slots_per_page = PAGE_SIZE >> reg->hr_block_bits;
1718 reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
1719
1720 mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
1721 reg->hr_start_block, reg->hr_blocks);
1722 mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
1723 reg->hr_block_bytes, reg->hr_block_bits);
1724 mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
1725 mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
1726 }
1727
1728 static int o2hb_map_slot_data(struct o2hb_region *reg)
1729 {
1730 int i, j;
1731 unsigned int last_slot;
1732 unsigned int spp = reg->hr_slots_per_page;
1733 struct page *page;
1734 char *raw;
1735 struct o2hb_disk_slot *slot;
1736
1737 reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
1738 if (reg->hr_tmp_block == NULL)
1739 return -ENOMEM;
1740
1741 reg->hr_slots = kcalloc(reg->hr_blocks,
1742 sizeof(struct o2hb_disk_slot), GFP_KERNEL);
1743 if (reg->hr_slots == NULL)
1744 return -ENOMEM;
1745
1746 for(i = 0; i < reg->hr_blocks; i++) {
1747 slot = ®->hr_slots[i];
1748 slot->ds_node_num = i;
1749 INIT_LIST_HEAD(&slot->ds_live_item);
1750 slot->ds_raw_block = NULL;
1751 }
1752
1753 reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
1754 mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
1755 "at %u blocks per page\n",
1756 reg->hr_num_pages, reg->hr_blocks, spp);
1757
1758 reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
1759 GFP_KERNEL);
1760 if (!reg->hr_slot_data)
1761 return -ENOMEM;
1762
1763 for(i = 0; i < reg->hr_num_pages; i++) {
1764 page = alloc_page(GFP_KERNEL);
1765 if (!page)
1766 return -ENOMEM;
1767
1768 reg->hr_slot_data[i] = page;
1769
1770 last_slot = i * spp;
1771 raw = page_address(page);
1772 for (j = 0;
1773 (j < spp) && ((j + last_slot) < reg->hr_blocks);
1774 j++) {
1775 BUG_ON((j + last_slot) >= reg->hr_blocks);
1776
1777 slot = ®->hr_slots[j + last_slot];
1778 slot->ds_raw_block =
1779 (struct o2hb_disk_heartbeat_block *) raw;
1780
1781 raw += reg->hr_block_bytes;
1782 }
1783 }
1784
1785 return 0;
1786 }
1787
1788 /* Read in all the slots available and populate the tracking
1789 * structures so that we can start with a baseline idea of what's
1790 * there. */
1791 static int o2hb_populate_slot_data(struct o2hb_region *reg)
1792 {
1793 int ret, i;
1794 struct o2hb_disk_slot *slot;
1795 struct o2hb_disk_heartbeat_block *hb_block;
1796
1797 ret = o2hb_read_slots(reg, 0, reg->hr_blocks);
1798 if (ret)
1799 goto out;
1800
1801 /* We only want to get an idea of the values initially in each
1802 * slot, so we do no verification - o2hb_check_slot will
1803 * actually determine if each configured slot is valid and
1804 * whether any values have changed. */
1805 for(i = 0; i < reg->hr_blocks; i++) {
1806 slot = ®->hr_slots[i];
1807 hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
1808
1809 /* Only fill the values that o2hb_check_slot uses to
1810 * determine changing slots */
1811 slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
1812 slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
1813 }
1814
1815 out:
1816 return ret;
1817 }
1818
1819 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
1820 static ssize_t o2hb_region_dev_store(struct config_item *item,
1821 const char *page,
1822 size_t count)
1823 {
1824 struct o2hb_region *reg = to_o2hb_region(item);
1825 struct task_struct *hb_task;
1826 long fd;
1827 int sectsize;
1828 char *p = (char *)page;
1829 struct fd f;
1830 struct inode *inode;
1831 ssize_t ret = -EINVAL;
1832 int live_threshold;
1833
1834 if (reg->hr_bdev)
1835 goto out;
1836
1837 /* We can't heartbeat without having had our node number
1838 * configured yet. */
1839 if (o2nm_this_node() == O2NM_MAX_NODES)
1840 goto out;
1841
1842 fd = simple_strtol(p, &p, 0);
1843 if (!p || (*p && (*p != '\n')))
1844 goto out;
1845
1846 if (fd < 0 || fd >= INT_MAX)
1847 goto out;
1848
1849 f = fdget(fd);
1850 if (f.file == NULL)
1851 goto out;
1852
1853 if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
1854 reg->hr_block_bytes == 0)
1855 goto out2;
1856
1857 inode = igrab(f.file->f_mapping->host);
1858 if (inode == NULL)
1859 goto out2;
1860
1861 if (!S_ISBLK(inode->i_mode))
1862 goto out3;
1863
1864 reg->hr_bdev = I_BDEV(f.file->f_mapping->host);
1865 ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL);
1866 if (ret) {
1867 reg->hr_bdev = NULL;
1868 goto out3;
1869 }
1870 inode = NULL;
1871
1872 bdevname(reg->hr_bdev, reg->hr_dev_name);
1873
1874 sectsize = bdev_logical_block_size(reg->hr_bdev);
1875 if (sectsize != reg->hr_block_bytes) {
1876 mlog(ML_ERROR,
1877 "blocksize %u incorrect for device, expected %d",
1878 reg->hr_block_bytes, sectsize);
1879 ret = -EINVAL;
1880 goto out3;
1881 }
1882
1883 o2hb_init_region_params(reg);
1884
1885 /* Generation of zero is invalid */
1886 do {
1887 get_random_bytes(®->hr_generation,
1888 sizeof(reg->hr_generation));
1889 } while (reg->hr_generation == 0);
1890
1891 ret = o2hb_map_slot_data(reg);
1892 if (ret) {
1893 mlog_errno(ret);
1894 goto out3;
1895 }
1896
1897 ret = o2hb_populate_slot_data(reg);
1898 if (ret) {
1899 mlog_errno(ret);
1900 goto out3;
1901 }
1902
1903 INIT_DELAYED_WORK(®->hr_write_timeout_work, o2hb_write_timeout);
1904 INIT_DELAYED_WORK(®->hr_nego_timeout_work, o2hb_nego_timeout);
1905
1906 /*
1907 * A node is considered live after it has beat LIVE_THRESHOLD
1908 * times. We're not steady until we've given them a chance
1909 * _after_ our first read.
1910 * The default threshold is bare minimum so as to limit the delay
1911 * during mounts. For global heartbeat, the threshold doubled for the
1912 * first region.
1913 */
1914 live_threshold = O2HB_LIVE_THRESHOLD;
1915 if (o2hb_global_heartbeat_active()) {
1916 spin_lock(&o2hb_live_lock);
1917 if (bitmap_weight(o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1)
1918 live_threshold <<= 1;
1919 spin_unlock(&o2hb_live_lock);
1920 }
1921 ++live_threshold;
1922 atomic_set(®->hr_steady_iterations, live_threshold);
1923 /* unsteady_iterations is triple the steady_iterations */
1924 atomic_set(®->hr_unsteady_iterations, (live_threshold * 3));
1925
1926 hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
1927 reg->hr_item.ci_name);
1928 if (IS_ERR(hb_task)) {
1929 ret = PTR_ERR(hb_task);
1930 mlog_errno(ret);
1931 goto out3;
1932 }
1933
1934 spin_lock(&o2hb_live_lock);
1935 reg->hr_task = hb_task;
1936 spin_unlock(&o2hb_live_lock);
1937
1938 ret = wait_event_interruptible(o2hb_steady_queue,
1939 atomic_read(®->hr_steady_iterations) == 0 ||
1940 reg->hr_node_deleted);
1941 if (ret) {
1942 atomic_set(®->hr_steady_iterations, 0);
1943 reg->hr_aborted_start = 1;
1944 }
1945
1946 if (reg->hr_aborted_start) {
1947 ret = -EIO;
1948 goto out3;
1949 }
1950
1951 if (reg->hr_node_deleted) {
1952 ret = -EINVAL;
1953 goto out3;
1954 }
1955
1956 /* Ok, we were woken. Make sure it wasn't by drop_item() */
1957 spin_lock(&o2hb_live_lock);
1958 hb_task = reg->hr_task;
1959 if (o2hb_global_heartbeat_active())
1960 set_bit(reg->hr_region_num, o2hb_live_region_bitmap);
1961 spin_unlock(&o2hb_live_lock);
1962
1963 if (hb_task)
1964 ret = count;
1965 else
1966 ret = -EIO;
1967
1968 if (hb_task && o2hb_global_heartbeat_active())
1969 printk(KERN_NOTICE "o2hb: Heartbeat started on region %s (%s)\n",
1970 config_item_name(®->hr_item), reg->hr_dev_name);
1971
1972 out3:
1973 iput(inode);
1974 out2:
1975 fdput(f);
1976 out:
1977 if (ret < 0) {
1978 if (reg->hr_bdev) {
1979 blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1980 reg->hr_bdev = NULL;
1981 }
1982 }
1983 return ret;
1984 }
1985
1986 static ssize_t o2hb_region_pid_show(struct config_item *item, char *page)
1987 {
1988 struct o2hb_region *reg = to_o2hb_region(item);
1989 pid_t pid = 0;
1990
1991 spin_lock(&o2hb_live_lock);
1992 if (reg->hr_task)
1993 pid = task_pid_nr(reg->hr_task);
1994 spin_unlock(&o2hb_live_lock);
1995
1996 if (!pid)
1997 return 0;
1998
1999 return sprintf(page, "%u\n", pid);
2000 }
2001
2002 CONFIGFS_ATTR(o2hb_region_, block_bytes);
2003 CONFIGFS_ATTR(o2hb_region_, start_block);
2004 CONFIGFS_ATTR(o2hb_region_, blocks);
2005 CONFIGFS_ATTR(o2hb_region_, dev);
2006 CONFIGFS_ATTR_RO(o2hb_region_, pid);
2007
2008 static struct configfs_attribute *o2hb_region_attrs[] = {
2009 &o2hb_region_attr_block_bytes,
2010 &o2hb_region_attr_start_block,
2011 &o2hb_region_attr_blocks,
2012 &o2hb_region_attr_dev,
2013 &o2hb_region_attr_pid,
2014 NULL,
2015 };
2016
2017 static struct configfs_item_operations o2hb_region_item_ops = {
2018 .release = o2hb_region_release,
2019 };
2020
2021 static const struct config_item_type o2hb_region_type = {
2022 .ct_item_ops = &o2hb_region_item_ops,
2023 .ct_attrs = o2hb_region_attrs,
2024 .ct_owner = THIS_MODULE,
2025 };
2026
2027 /* heartbeat set */
2028
2029 struct o2hb_heartbeat_group {
2030 struct config_group hs_group;
2031 /* some stuff? */
2032 };
2033
2034 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
2035 {
2036 return group ?
2037 container_of(group, struct o2hb_heartbeat_group, hs_group)
2038 : NULL;
2039 }
2040
2041 static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir)
2042 {
2043 int ret = -ENOMEM;
2044
2045 reg->hr_debug_dir =
2046 debugfs_create_dir(config_item_name(®->hr_item), dir);
2047 if (!reg->hr_debug_dir) {
2048 mlog_errno(ret);
2049 goto bail;
2050 }
2051
2052 reg->hr_debug_livenodes =
2053 o2hb_debug_create(O2HB_DEBUG_LIVENODES,
2054 reg->hr_debug_dir,
2055 &(reg->hr_db_livenodes),
2056 sizeof(*(reg->hr_db_livenodes)),
2057 O2HB_DB_TYPE_REGION_LIVENODES,
2058 sizeof(reg->hr_live_node_bitmap),
2059 O2NM_MAX_NODES, reg);
2060 if (!reg->hr_debug_livenodes) {
2061 mlog_errno(ret);
2062 goto bail;
2063 }
2064
2065 reg->hr_debug_regnum =
2066 o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER,
2067 reg->hr_debug_dir,
2068 &(reg->hr_db_regnum),
2069 sizeof(*(reg->hr_db_regnum)),
2070 O2HB_DB_TYPE_REGION_NUMBER,
2071 0, O2NM_MAX_NODES, reg);
2072 if (!reg->hr_debug_regnum) {
2073 mlog_errno(ret);
2074 goto bail;
2075 }
2076
2077 reg->hr_debug_elapsed_time =
2078 o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME,
2079 reg->hr_debug_dir,
2080 &(reg->hr_db_elapsed_time),
2081 sizeof(*(reg->hr_db_elapsed_time)),
2082 O2HB_DB_TYPE_REGION_ELAPSED_TIME,
2083 0, 0, reg);
2084 if (!reg->hr_debug_elapsed_time) {
2085 mlog_errno(ret);
2086 goto bail;
2087 }
2088
2089 reg->hr_debug_pinned =
2090 o2hb_debug_create(O2HB_DEBUG_REGION_PINNED,
2091 reg->hr_debug_dir,
2092 &(reg->hr_db_pinned),
2093 sizeof(*(reg->hr_db_pinned)),
2094 O2HB_DB_TYPE_REGION_PINNED,
2095 0, 0, reg);
2096 if (!reg->hr_debug_pinned) {
2097 mlog_errno(ret);
2098 goto bail;
2099 }
2100
2101 ret = 0;
2102 bail:
2103 return ret;
2104 }
2105
2106 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
2107 const char *name)
2108 {
2109 struct o2hb_region *reg = NULL;
2110 int ret;
2111
2112 reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
2113 if (reg == NULL)
2114 return ERR_PTR(-ENOMEM);
2115
2116 if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) {
2117 ret = -ENAMETOOLONG;
2118 goto free;
2119 }
2120
2121 spin_lock(&o2hb_live_lock);
2122 reg->hr_region_num = 0;
2123 if (o2hb_global_heartbeat_active()) {
2124 reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap,
2125 O2NM_MAX_REGIONS);
2126 if (reg->hr_region_num >= O2NM_MAX_REGIONS) {
2127 spin_unlock(&o2hb_live_lock);
2128 ret = -EFBIG;
2129 goto free;
2130 }
2131 set_bit(reg->hr_region_num, o2hb_region_bitmap);
2132 }
2133 list_add_tail(®->hr_all_item, &o2hb_all_regions);
2134 spin_unlock(&o2hb_live_lock);
2135
2136 config_item_init_type_name(®->hr_item, name, &o2hb_region_type);
2137
2138 /* this is the same way to generate msg key as dlm, for local heartbeat,
2139 * name is also the same, so make initial crc value different to avoid
2140 * message key conflict.
2141 */
2142 reg->hr_key = crc32_le(reg->hr_region_num + O2NM_MAX_REGIONS,
2143 name, strlen(name));
2144 INIT_LIST_HEAD(®->hr_handler_list);
2145 ret = o2net_register_handler(O2HB_NEGO_TIMEOUT_MSG, reg->hr_key,
2146 sizeof(struct o2hb_nego_msg),
2147 o2hb_nego_timeout_handler,
2148 reg, NULL, ®->hr_handler_list);
2149 if (ret)
2150 goto free;
2151
2152 ret = o2net_register_handler(O2HB_NEGO_APPROVE_MSG, reg->hr_key,
2153 sizeof(struct o2hb_nego_msg),
2154 o2hb_nego_approve_handler,
2155 reg, NULL, ®->hr_handler_list);
2156 if (ret)
2157 goto unregister_handler;
2158
2159 ret = o2hb_debug_region_init(reg, o2hb_debug_dir);
2160 if (ret) {
2161 config_item_put(®->hr_item);
2162 goto unregister_handler;
2163 }
2164
2165 return ®->hr_item;
2166
2167 unregister_handler:
2168 o2net_unregister_handler_list(®->hr_handler_list);
2169 free:
2170 kfree(reg);
2171 return ERR_PTR(ret);
2172 }
2173
2174 static void o2hb_heartbeat_group_drop_item(struct config_group *group,
2175 struct config_item *item)
2176 {
2177 struct task_struct *hb_task;
2178 struct o2hb_region *reg = to_o2hb_region(item);
2179 int quorum_region = 0;
2180
2181 /* stop the thread when the user removes the region dir */
2182 spin_lock(&o2hb_live_lock);
2183 hb_task = reg->hr_task;
2184 reg->hr_task = NULL;
2185 reg->hr_item_dropped = 1;
2186 spin_unlock(&o2hb_live_lock);
2187
2188 if (hb_task)
2189 kthread_stop(hb_task);
2190
2191 if (o2hb_global_heartbeat_active()) {
2192 spin_lock(&o2hb_live_lock);
2193 clear_bit(reg->hr_region_num, o2hb_region_bitmap);
2194 clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
2195 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
2196 quorum_region = 1;
2197 clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
2198 spin_unlock(&o2hb_live_lock);
2199 printk(KERN_NOTICE "o2hb: Heartbeat %s on region %s (%s)\n",
2200 ((atomic_read(®->hr_steady_iterations) == 0) ?
2201 "stopped" : "start aborted"), config_item_name(item),
2202 reg->hr_dev_name);
2203 }
2204
2205 /*
2206 * If we're racing a dev_write(), we need to wake them. They will
2207 * check reg->hr_task
2208 */
2209 if (atomic_read(®->hr_steady_iterations) != 0) {
2210 reg->hr_aborted_start = 1;
2211 atomic_set(®->hr_steady_iterations, 0);
2212 wake_up(&o2hb_steady_queue);
2213 }
2214
2215 config_item_put(item);
2216
2217 if (!o2hb_global_heartbeat_active() || !quorum_region)
2218 return;
2219
2220 /*
2221 * If global heartbeat active and there are dependent users,
2222 * pin all regions if quorum region count <= CUT_OFF
2223 */
2224 spin_lock(&o2hb_live_lock);
2225
2226 if (!o2hb_dependent_users)
2227 goto unlock;
2228
2229 if (bitmap_weight(o2hb_quorum_region_bitmap,
2230 O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2231 o2hb_region_pin(NULL);
2232
2233 unlock:
2234 spin_unlock(&o2hb_live_lock);
2235 }
2236
2237 static ssize_t o2hb_heartbeat_group_dead_threshold_show(struct config_item *item,
2238 char *page)
2239 {
2240 return sprintf(page, "%u\n", o2hb_dead_threshold);
2241 }
2242
2243 static ssize_t o2hb_heartbeat_group_dead_threshold_store(struct config_item *item,
2244 const char *page, size_t count)
2245 {
2246 unsigned long tmp;
2247 char *p = (char *)page;
2248
2249 tmp = simple_strtoul(p, &p, 10);
2250 if (!p || (*p && (*p != '\n')))
2251 return -EINVAL;
2252
2253 /* this will validate ranges for us. */
2254 o2hb_dead_threshold_set((unsigned int) tmp);
2255
2256 return count;
2257 }
2258
2259 static ssize_t o2hb_heartbeat_group_mode_show(struct config_item *item,
2260 char *page)
2261 {
2262 return sprintf(page, "%s\n",
2263 o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]);
2264 }
2265
2266 static ssize_t o2hb_heartbeat_group_mode_store(struct config_item *item,
2267 const char *page, size_t count)
2268 {
2269 unsigned int i;
2270 int ret;
2271 size_t len;
2272
2273 len = (page[count - 1] == '\n') ? count - 1 : count;
2274 if (!len)
2275 return -EINVAL;
2276
2277 for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) {
2278 if (strncasecmp(page, o2hb_heartbeat_mode_desc[i], len))
2279 continue;
2280
2281 ret = o2hb_global_heartbeat_mode_set(i);
2282 if (!ret)
2283 printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n",
2284 o2hb_heartbeat_mode_desc[i]);
2285 return count;
2286 }
2287
2288 return -EINVAL;
2289
2290 }
2291
2292 CONFIGFS_ATTR(o2hb_heartbeat_group_, dead_threshold);
2293 CONFIGFS_ATTR(o2hb_heartbeat_group_, mode);
2294
2295 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
2296 &o2hb_heartbeat_group_attr_dead_threshold,
2297 &o2hb_heartbeat_group_attr_mode,
2298 NULL,
2299 };
2300
2301 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
2302 .make_item = o2hb_heartbeat_group_make_item,
2303 .drop_item = o2hb_heartbeat_group_drop_item,
2304 };
2305
2306 static const struct config_item_type o2hb_heartbeat_group_type = {
2307 .ct_group_ops = &o2hb_heartbeat_group_group_ops,
2308 .ct_attrs = o2hb_heartbeat_group_attrs,
2309 .ct_owner = THIS_MODULE,
2310 };
2311
2312 /* this is just here to avoid touching group in heartbeat.h which the
2313 * entire damn world #includes */
2314 struct config_group *o2hb_alloc_hb_set(void)
2315 {
2316 struct o2hb_heartbeat_group *hs = NULL;
2317 struct config_group *ret = NULL;
2318
2319 hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
2320 if (hs == NULL)
2321 goto out;
2322
2323 config_group_init_type_name(&hs->hs_group, "heartbeat",
2324 &o2hb_heartbeat_group_type);
2325
2326 ret = &hs->hs_group;
2327 out:
2328 if (ret == NULL)
2329 kfree(hs);
2330 return ret;
2331 }
2332
2333 void o2hb_free_hb_set(struct config_group *group)
2334 {
2335 struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
2336 kfree(hs);
2337 }
2338
2339 /* hb callback registration and issuing */
2340
2341 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
2342 {
2343 if (type == O2HB_NUM_CB)
2344 return ERR_PTR(-EINVAL);
2345
2346 return &o2hb_callbacks[type];
2347 }
2348
2349 void o2hb_setup_callback(struct o2hb_callback_func *hc,
2350 enum o2hb_callback_type type,
2351 o2hb_cb_func *func,
2352 void *data,
2353 int priority)
2354 {
2355 INIT_LIST_HEAD(&hc->hc_item);
2356 hc->hc_func = func;
2357 hc->hc_data = data;
2358 hc->hc_priority = priority;
2359 hc->hc_type = type;
2360 hc->hc_magic = O2HB_CB_MAGIC;
2361 }
2362 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
2363
2364 /*
2365 * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2366 * In global heartbeat mode, region_uuid passed is NULL.
2367 *
2368 * In local, we only pin the matching region. In global we pin all the active
2369 * regions.
2370 */
2371 static int o2hb_region_pin(const char *region_uuid)
2372 {
2373 int ret = 0, found = 0;
2374 struct o2hb_region *reg;
2375 char *uuid;
2376
2377 assert_spin_locked(&o2hb_live_lock);
2378
2379 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2380 if (reg->hr_item_dropped)
2381 continue;
2382
2383 uuid = config_item_name(®->hr_item);
2384
2385 /* local heartbeat */
2386 if (region_uuid) {
2387 if (strcmp(region_uuid, uuid))
2388 continue;
2389 found = 1;
2390 }
2391
2392 if (reg->hr_item_pinned || reg->hr_item_dropped)
2393 goto skip_pin;
2394
2395 /* Ignore ENOENT only for local hb (userdlm domain) */
2396 ret = o2nm_depend_item(®->hr_item);
2397 if (!ret) {
2398 mlog(ML_CLUSTER, "Pin region %s\n", uuid);
2399 reg->hr_item_pinned = 1;
2400 } else {
2401 if (ret == -ENOENT && found)
2402 ret = 0;
2403 else {
2404 mlog(ML_ERROR, "Pin region %s fails with %d\n",
2405 uuid, ret);
2406 break;
2407 }
2408 }
2409 skip_pin:
2410 if (found)
2411 break;
2412 }
2413
2414 return ret;
2415 }
2416
2417 /*
2418 * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2419 * In global heartbeat mode, region_uuid passed is NULL.
2420 *
2421 * In local, we only unpin the matching region. In global we unpin all the
2422 * active regions.
2423 */
2424 static void o2hb_region_unpin(const char *region_uuid)
2425 {
2426 struct o2hb_region *reg;
2427 char *uuid;
2428 int found = 0;
2429
2430 assert_spin_locked(&o2hb_live_lock);
2431
2432 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2433 if (reg->hr_item_dropped)
2434 continue;
2435
2436 uuid = config_item_name(®->hr_item);
2437 if (region_uuid) {
2438 if (strcmp(region_uuid, uuid))
2439 continue;
2440 found = 1;
2441 }
2442
2443 if (reg->hr_item_pinned) {
2444 mlog(ML_CLUSTER, "Unpin region %s\n", uuid);
2445 o2nm_undepend_item(®->hr_item);
2446 reg->hr_item_pinned = 0;
2447 }
2448 if (found)
2449 break;
2450 }
2451 }
2452
2453 static int o2hb_region_inc_user(const char *region_uuid)
2454 {
2455 int ret = 0;
2456
2457 spin_lock(&o2hb_live_lock);
2458
2459 /* local heartbeat */
2460 if (!o2hb_global_heartbeat_active()) {
2461 ret = o2hb_region_pin(region_uuid);
2462 goto unlock;
2463 }
2464
2465 /*
2466 * if global heartbeat active and this is the first dependent user,
2467 * pin all regions if quorum region count <= CUT_OFF
2468 */
2469 o2hb_dependent_users++;
2470 if (o2hb_dependent_users > 1)
2471 goto unlock;
2472
2473 if (bitmap_weight(o2hb_quorum_region_bitmap,
2474 O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2475 ret = o2hb_region_pin(NULL);
2476
2477 unlock:
2478 spin_unlock(&o2hb_live_lock);
2479 return ret;
2480 }
2481
2482 static void o2hb_region_dec_user(const char *region_uuid)
2483 {
2484 spin_lock(&o2hb_live_lock);
2485
2486 /* local heartbeat */
2487 if (!o2hb_global_heartbeat_active()) {
2488 o2hb_region_unpin(region_uuid);
2489 goto unlock;
2490 }
2491
2492 /*
2493 * if global heartbeat active and there are no dependent users,
2494 * unpin all quorum regions
2495 */
2496 o2hb_dependent_users--;
2497 if (!o2hb_dependent_users)
2498 o2hb_region_unpin(NULL);
2499
2500 unlock:
2501 spin_unlock(&o2hb_live_lock);
2502 }
2503
2504 int o2hb_register_callback(const char *region_uuid,
2505 struct o2hb_callback_func *hc)
2506 {
2507 struct o2hb_callback_func *f;
2508 struct o2hb_callback *hbcall;
2509 int ret;
2510
2511 BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2512 BUG_ON(!list_empty(&hc->hc_item));
2513
2514 hbcall = hbcall_from_type(hc->hc_type);
2515 if (IS_ERR(hbcall)) {
2516 ret = PTR_ERR(hbcall);
2517 goto out;
2518 }
2519
2520 if (region_uuid) {
2521 ret = o2hb_region_inc_user(region_uuid);
2522 if (ret) {
2523 mlog_errno(ret);
2524 goto out;
2525 }
2526 }
2527
2528 down_write(&o2hb_callback_sem);
2529
2530 list_for_each_entry(f, &hbcall->list, hc_item) {
2531 if (hc->hc_priority < f->hc_priority) {
2532 list_add_tail(&hc->hc_item, &f->hc_item);
2533 break;
2534 }
2535 }
2536 if (list_empty(&hc->hc_item))
2537 list_add_tail(&hc->hc_item, &hbcall->list);
2538
2539 up_write(&o2hb_callback_sem);
2540 ret = 0;
2541 out:
2542 mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n",
2543 ret, __builtin_return_address(0), hc);
2544 return ret;
2545 }
2546 EXPORT_SYMBOL_GPL(o2hb_register_callback);
2547
2548 void o2hb_unregister_callback(const char *region_uuid,
2549 struct o2hb_callback_func *hc)
2550 {
2551 BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2552
2553 mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n",
2554 __builtin_return_address(0), hc);
2555
2556 /* XXX Can this happen _with_ a region reference? */
2557 if (list_empty(&hc->hc_item))
2558 return;
2559
2560 if (region_uuid)
2561 o2hb_region_dec_user(region_uuid);
2562
2563 down_write(&o2hb_callback_sem);
2564
2565 list_del_init(&hc->hc_item);
2566
2567 up_write(&o2hb_callback_sem);
2568 }
2569 EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
2570
2571 int o2hb_check_node_heartbeating_no_sem(u8 node_num)
2572 {
2573 unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2574
2575 spin_lock(&o2hb_live_lock);
2576 o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2577 spin_unlock(&o2hb_live_lock);
2578 if (!test_bit(node_num, testing_map)) {
2579 mlog(ML_HEARTBEAT,
2580 "node (%u) does not have heartbeating enabled.\n",
2581 node_num);
2582 return 0;
2583 }
2584
2585 return 1;
2586 }
2587 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_no_sem);
2588
2589 int o2hb_check_node_heartbeating_from_callback(u8 node_num)
2590 {
2591 unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2592
2593 o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2594 if (!test_bit(node_num, testing_map)) {
2595 mlog(ML_HEARTBEAT,
2596 "node (%u) does not have heartbeating enabled.\n",
2597 node_num);
2598 return 0;
2599 }
2600
2601 return 1;
2602 }
2603 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
2604
2605 /*
2606 * this is just a hack until we get the plumbing which flips file systems
2607 * read only and drops the hb ref instead of killing the node dead.
2608 */
2609 void o2hb_stop_all_regions(void)
2610 {
2611 struct o2hb_region *reg;
2612
2613 mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
2614
2615 spin_lock(&o2hb_live_lock);
2616
2617 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
2618 reg->hr_unclean_stop = 1;
2619
2620 spin_unlock(&o2hb_live_lock);
2621 }
2622 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
2623
2624 int o2hb_get_all_regions(char *region_uuids, u8 max_regions)
2625 {
2626 struct o2hb_region *reg;
2627 int numregs = 0;
2628 char *p;
2629
2630 spin_lock(&o2hb_live_lock);
2631
2632 p = region_uuids;
2633 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2634 if (reg->hr_item_dropped)
2635 continue;
2636
2637 mlog(0, "Region: %s\n", config_item_name(®->hr_item));
2638 if (numregs < max_regions) {
2639 memcpy(p, config_item_name(®->hr_item),
2640 O2HB_MAX_REGION_NAME_LEN);
2641 p += O2HB_MAX_REGION_NAME_LEN;
2642 }
2643 numregs++;
2644 }
2645
2646 spin_unlock(&o2hb_live_lock);
2647
2648 return numregs;
2649 }
2650 EXPORT_SYMBOL_GPL(o2hb_get_all_regions);
2651
2652 int o2hb_global_heartbeat_active(void)
2653 {
2654 return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL);
2655 }
2656 EXPORT_SYMBOL(o2hb_global_heartbeat_active);
2657
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.