aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorMark Fasheh <mark.fasheh@oracle.com>2005-12-15 17:31:23 -0500
committerJoel Becker <joel.becker@oracle.com>2006-01-03 14:45:46 -0500
commita7f6a5fb4bde142b622706e2006ba33f793e13ed (patch)
tree27db335992f61a13658e604ee56f4c6a3c79aea5
parent0c83ed8eeb28a045cdbd0b216679938aa9e665fe (diff)
[PATCH] OCFS2: The Second Oracle Cluster Filesystem
Disk based heartbeat. Configured and started from userspace, the kernel component handles I/O submission and event generation via callback mechanism. Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com> Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>
Diffstat
-rw-r--r--fs/ocfs2/cluster/heartbeat.c1797
-rw-r--r--fs/ocfs2/cluster/heartbeat.h82
-rw-r--r--fs/ocfs2/cluster/ocfs2_heartbeat.h37
3 files changed, 1916 insertions, 0 deletions
diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
new file mode 100644
index 000000000000..7307ba528913
--- /dev/null
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -0,0 +1,1797 @@
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 */
51static 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 */
57static spinlock_t o2hb_live_lock = SPIN_LOCK_UNLOCKED;
58static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
59static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
60static LIST_HEAD(o2hb_node_events);
61static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
62
63static LIST_HEAD(o2hb_all_regions);
64
65static struct o2hb_callback {
66 struct list_head list;
67} o2hb_callbacks[O2HB_NUM_CB];
68
69static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
70
71#define O2HB_DEFAULT_BLOCK_BITS 9
72
73unsigned 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. */
80static 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
90struct 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
97struct 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 */
109struct 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
153struct o2hb_bio_wait_ctxt {
154 atomic_t wc_num_reqs;
155 struct completion wc_io_complete;
156};
157
158static 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
168static 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
178static 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
184static 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 */
192static 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
205static 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
215static 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. */
233static 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
293bail:
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 */
303static 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
327static 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
348static 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
391bail_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
404static 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;
427bail:
428 return status;
429}
430
431static 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
450static 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
459static 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. */
472static 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
493static 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
523static 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 */
538static 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
584static 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
601static 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
632static 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
706fire_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 }
763out:
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. */
774static 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
791static 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. */
859static 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
878static 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 */
893static 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
956void 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 */
972static 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 */
983void 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}
993EXPORT_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
1000static 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. */
1008static 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
1039static 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
1066static 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
1072static 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
1094static 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
1100static 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
1119static 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
1125static 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
1147static 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
1158static 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
1171static 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. */
1242static 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
1270out:
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 */
1276static 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;
1383out:
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
1397struct 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
1403static 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
1411static 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
1419static 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
1427static 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
1435static 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
1443static 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
1457static 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
1471static 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
1477static 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
1485struct o2hb_heartbeat_group {
1486 struct config_group hs_group;
1487 /* some stuff? */
1488};
1489
1490static 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
1497static 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);
1514out:
1515 if (ret == NULL)
1516 kfree(reg);
1517
1518 return ret;
1519}
1520
1521static 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
1535struct 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
1541static 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
1555static 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
1569static 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
1575static 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
1592static 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
1600static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
1601 &o2hb_heartbeat_group_attr_threshold.attr,
1602 NULL,
1603};
1604
1605static 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
1610static 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
1615static 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 */
1624struct 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;
1637out:
1638 if (ret == NULL)
1639 kfree(hs);
1640 return ret;
1641}
1642
1643void 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
1651static 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
1659void 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}
1672EXPORT_SYMBOL_GPL(o2hb_setup_callback);
1673
1674int 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;
1704out:
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}
1709EXPORT_SYMBOL_GPL(o2hb_register_callback);
1710
1711int 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}
1729EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
1730
1731int 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}
1745EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating);
1746
1747int 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}
1761EXPORT_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. */
1765int 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}
1778EXPORT_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 */
1784void 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}
1797EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
diff --git a/fs/ocfs2/cluster/heartbeat.h b/fs/ocfs2/cluster/heartbeat.h
new file mode 100644
index 000000000000..cac6223206a9
--- /dev/null
+++ b/fs/ocfs2/cluster/heartbeat.h
@@ -0,0 +1,82 @@
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * heartbeat.h
5 *
6 * Function prototypes
7 *
8 * Copyright (C) 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 *
25 */
26
27#ifndef O2CLUSTER_HEARTBEAT_H
28#define O2CLUSTER_HEARTBEAT_H
29
30#include "ocfs2_heartbeat.h"
31
32#define O2HB_REGION_TIMEOUT_MS 2000
33
34/* number of changes to be seen as live */
35#define O2HB_LIVE_THRESHOLD 2
36/* number of equal samples to be seen as dead */
37extern unsigned int o2hb_dead_threshold;
38#define O2HB_DEFAULT_DEAD_THRESHOLD 7
39/* Otherwise MAX_WRITE_TIMEOUT will be zero... */
40#define O2HB_MIN_DEAD_THRESHOLD 2
41#define O2HB_MAX_WRITE_TIMEOUT_MS (O2HB_REGION_TIMEOUT_MS * (o2hb_dead_threshold - 1))
42
43#define O2HB_CB_MAGIC 0x51d1e4ec
44
45/* callback stuff */
46enum o2hb_callback_type {
47 O2HB_NODE_DOWN_CB = 0,
48 O2HB_NODE_UP_CB,
49 O2HB_NUM_CB
50};
51
52struct o2nm_node;
53typedef void (o2hb_cb_func)(struct o2nm_node *, int, void *);
54
55struct o2hb_callback_func {
56 u32 hc_magic;
57 struct list_head hc_item;
58 o2hb_cb_func *hc_func;
59 void *hc_data;
60 int hc_priority;
61 enum o2hb_callback_type hc_type;
62};
63
64struct config_group *o2hb_alloc_hb_set(void);
65void o2hb_free_hb_set(struct config_group *group);
66
67void o2hb_setup_callback(struct o2hb_callback_func *hc,
68 enum o2hb_callback_type type,
69 o2hb_cb_func *func,
70 void *data,
71 int priority);
72int o2hb_register_callback(struct o2hb_callback_func *hc);
73int o2hb_unregister_callback(struct o2hb_callback_func *hc);
74void o2hb_fill_node_map(unsigned long *map,
75 unsigned bytes);
76void o2hb_init(void);
77int o2hb_check_node_heartbeating(u8 node_num);
78int o2hb_check_node_heartbeating_from_callback(u8 node_num);
79int o2hb_check_local_node_heartbeating(void);
80void o2hb_stop_all_regions(void);
81
82#endif /* O2CLUSTER_HEARTBEAT_H */
diff --git a/fs/ocfs2/cluster/ocfs2_heartbeat.h b/fs/ocfs2/cluster/ocfs2_heartbeat.h
new file mode 100644
index 000000000000..94096069cb43
--- /dev/null
+++ b/fs/ocfs2/cluster/ocfs2_heartbeat.h
@@ -0,0 +1,37 @@
1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * ocfs2_heartbeat.h
5 *
6 * On-disk structures for ocfs2_heartbeat
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#ifndef _OCFS2_HEARTBEAT_H
27#define _OCFS2_HEARTBEAT_H
28
29struct o2hb_disk_heartbeat_block {
30 __le64 hb_seq;
31 __u8 hb_node;
32 __u8 hb_pad1[3];
33 __le32 hb_cksum;
34 __le64 hb_generation;
35};
36
37#endif /* _OCFS2_HEARTBEAT_H */
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-17 06:08:43 -0500