fs/ocfs2/cluster/heartbeat.c

   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 #include <linux/debugfs.h>
  37 #include <linux/slab.h>
  38
  39 #include "heartbeat.h"
  40 #include "tcp.h"
  41 #include "nodemanager.h"
  42 #include "quorum.h"
  43
  44 #include "masklog.h"
  45
  46
  47 /*
  48  * The first heartbeat pass had one global thread that would serialize all hb
  49  * callback calls.  This global serializing sem should only be removed once
  50  * we've made sure that all callees can deal with being called concurrently
  51  * from multiple hb region threads.
  52  */
  53 static DECLARE_RWSEM(o2hb_callback_sem);
  54
  55 /*
  56  * multiple hb threads are watching multiple regions.  A node is live
  57  * whenever any of the threads sees activity from the node in its region.
  58  */
  59 static DEFINE_SPINLOCK(o2hb_live_lock);
  60 static struct list_head o2hb_live_slots[O2NM_MAX_NODES];
  61 static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
  62 static LIST_HEAD(o2hb_node_events);
  63 static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue);
  64
  65 /*
  66  * In global heartbeat, we maintain a series of region bitmaps.
  67  *      - o2hb_region_bitmap allows us to limit the region number to max region.
  68  *      - o2hb_live_region_bitmap tracks live regions (seen steady iterations).
  69  *      - o2hb_quorum_region_bitmap tracks live regions that have seen all nodes
  70  *              heartbeat on it.
  71  *      - o2hb_failed_region_bitmap tracks the regions that have seen io timeouts.
  72  */
  73 static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
  74 static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
  75 static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
  76 static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)];
  77
  78 #define O2HB_DB_TYPE_LIVENODES          0
  79 #define O2HB_DB_TYPE_LIVEREGIONS        1
  80 #define O2HB_DB_TYPE_QUORUMREGIONS      2
  81 #define O2HB_DB_TYPE_FAILEDREGIONS      3
  82 #define O2HB_DB_TYPE_REGION_LIVENODES   4
  83 #define O2HB_DB_TYPE_REGION_NUMBER      5
  84 #define O2HB_DB_TYPE_REGION_ELAPSED_TIME        6
  85 #define O2HB_DB_TYPE_REGION_PINNED      7
  86 struct o2hb_debug_buf {
  87         int db_type;
  88         int db_size;
  89         int db_len;
  90         void *db_data;
  91 };
  92
  93 static struct o2hb_debug_buf *o2hb_db_livenodes;
  94 static struct o2hb_debug_buf *o2hb_db_liveregions;
  95 static struct o2hb_debug_buf *o2hb_db_quorumregions;
  96 static struct o2hb_debug_buf *o2hb_db_failedregions;
  97
  98 #define O2HB_DEBUG_DIR                  "o2hb"
  99 #define O2HB_DEBUG_LIVENODES            "livenodes"
 100 #define O2HB_DEBUG_LIVEREGIONS          "live_regions"
 101 #define O2HB_DEBUG_QUORUMREGIONS        "quorum_regions"
 102 #define O2HB_DEBUG_FAILEDREGIONS        "failed_regions"
 103 #define O2HB_DEBUG_REGION_NUMBER        "num"
 104 #define O2HB_DEBUG_REGION_ELAPSED_TIME  "elapsed_time_in_ms"
 105 #define O2HB_DEBUG_REGION_PINNED        "pinned"
 106
 107 static struct dentry *o2hb_debug_dir;
 108 static struct dentry *o2hb_debug_livenodes;
 109 static struct dentry *o2hb_debug_liveregions;
 110 static struct dentry *o2hb_debug_quorumregions;
 111 static struct dentry *o2hb_debug_failedregions;
 112
 113 static LIST_HEAD(o2hb_all_regions);
 114
 115 static struct o2hb_callback {
 116         struct list_head list;
 117 } o2hb_callbacks[O2HB_NUM_CB];
 118
 119 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type);
 120
 121 #define O2HB_DEFAULT_BLOCK_BITS       9
 122
 123 enum o2hb_heartbeat_modes {
 124         O2HB_HEARTBEAT_LOCAL            = 0,
 125         O2HB_HEARTBEAT_GLOBAL,
 126         O2HB_HEARTBEAT_NUM_MODES,
 127 };
 128
 129 char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = {
 130                 "local",        /* O2HB_HEARTBEAT_LOCAL */
 131                 "global",       /* O2HB_HEARTBEAT_GLOBAL */
 132 };
 133
 134 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD;
 135 unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL;
 136
 137 /*
 138  * o2hb_dependent_users tracks the number of registered callbacks that depend
 139  * on heartbeat. o2net and o2dlm are two entities that register this callback.
 140  * However only o2dlm depends on the heartbeat. It does not want the heartbeat
 141  * to stop while a dlm domain is still active.
 142  */
 143 unsigned int o2hb_dependent_users;
 144
 145 /*
 146  * In global heartbeat mode, all regions are pinned if there are one or more
 147  * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All
 148  * regions are unpinned if the region count exceeds the cut off or the number
 149  * of dependent users falls to zero.
 150  */
 151 #define O2HB_PIN_CUT_OFF                3
 152
 153 /*
 154  * In local heartbeat mode, we assume the dlm domain name to be the same as
 155  * region uuid. This is true for domains created for the file system but not
 156  * necessarily true for userdlm domains. This is a known limitation.
 157  *
 158  * In global heartbeat mode, we pin/unpin all o2hb regions. This solution
 159  * works for both file system and userdlm domains.
 160  */
 161 static int o2hb_region_pin(const char *region_uuid);
 162 static void o2hb_region_unpin(const char *region_uuid);
 163
 164 /* Only sets a new threshold if there are no active regions.
 165  *
 166  * No locking or otherwise interesting code is required for reading
 167  * o2hb_dead_threshold as it can't change once regions are active and
 168  * it's not interesting to anyone until then anyway. */
 169 static void o2hb_dead_threshold_set(unsigned int threshold)
 170 {
 171         if (threshold > O2HB_MIN_DEAD_THRESHOLD) {
 172                 spin_lock(&o2hb_live_lock);
 173                 if (list_empty(&o2hb_all_regions))
 174                         o2hb_dead_threshold = threshold;
 175                 spin_unlock(&o2hb_live_lock);
 176         }
 177 }
 178
 179 static int o2hb_global_hearbeat_mode_set(unsigned int hb_mode)
 180 {
 181         int ret = -1;
 182
 183         if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) {
 184                 spin_lock(&o2hb_live_lock);
 185                 if (list_empty(&o2hb_all_regions)) {
 186                         o2hb_heartbeat_mode = hb_mode;
 187                         ret = 0;
 188                 }
 189                 spin_unlock(&o2hb_live_lock);
 190         }
 191
 192         return ret;
 193 }
 194
 195 struct o2hb_node_event {
 196         struct list_head        hn_item;
 197         enum o2hb_callback_type hn_event_type;
 198         struct o2nm_node        *hn_node;
 199         int                     hn_node_num;
 200 };
 201
 202 struct o2hb_disk_slot {
 203         struct o2hb_disk_heartbeat_block *ds_raw_block;
 204         u8                      ds_node_num;
 205         u64                     ds_last_time;
 206         u64                     ds_last_generation;
 207         u16                     ds_equal_samples;
 208         u16                     ds_changed_samples;
 209         struct list_head        ds_live_item;
 210 };
 211
 212 /* each thread owns a region.. when we're asked to tear down the region
 213  * we ask the thread to stop, who cleans up the region */
 214 struct o2hb_region {
 215         struct config_item      hr_item;
 216
 217         struct list_head        hr_all_item;
 218         unsigned                hr_unclean_stop:1,
 219                                 hr_item_pinned:1,
 220                                 hr_item_dropped:1;
 221
 222         /* protected by the hr_callback_sem */
 223         struct task_struct      *hr_task;
 224
 225         unsigned int            hr_blocks;
 226         unsigned long long      hr_start_block;
 227
 228         unsigned int            hr_block_bits;
 229         unsigned int            hr_block_bytes;
 230
 231         unsigned int            hr_slots_per_page;
 232         unsigned int            hr_num_pages;
 233
 234         struct page             **hr_slot_data;
 235         struct block_device     *hr_bdev;
 236         struct o2hb_disk_slot   *hr_slots;
 237
 238         /* live node map of this region */
 239         unsigned long           hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
 240         unsigned int            hr_region_num;
 241
 242         struct dentry           *hr_debug_dir;
 243         struct dentry           *hr_debug_livenodes;
 244         struct dentry           *hr_debug_regnum;
 245         struct dentry           *hr_debug_elapsed_time;
 246         struct dentry           *hr_debug_pinned;
 247         struct o2hb_debug_buf   *hr_db_livenodes;
 248         struct o2hb_debug_buf   *hr_db_regnum;
 249         struct o2hb_debug_buf   *hr_db_elapsed_time;
 250         struct o2hb_debug_buf   *hr_db_pinned;
 251
 252         /* let the person setting up hb wait for it to return until it
 253          * has reached a 'steady' state.  This will be fixed when we have
 254          * a more complete api that doesn't lead to this sort of fragility. */
 255         atomic_t                hr_steady_iterations;
 256
 257         char                    hr_dev_name[BDEVNAME_SIZE];
 258
 259         unsigned int            hr_timeout_ms;
 260
 261         /* randomized as the region goes up and down so that a node
 262          * recognizes a node going up and down in one iteration */
 263         u64                     hr_generation;
 264
 265         struct delayed_work     hr_write_timeout_work;
 266         unsigned long           hr_last_timeout_start;
 267
 268         /* Used during o2hb_check_slot to hold a copy of the block
 269          * being checked because we temporarily have to zero out the
 270          * crc field. */
 271         struct o2hb_disk_heartbeat_block *hr_tmp_block;
 272 };
 273
 274 struct o2hb_bio_wait_ctxt {
 275         atomic_t          wc_num_reqs;
 276         struct completion wc_io_complete;
 277         int               wc_error;
 278 };
 279
 280 static int o2hb_pop_count(void *map, int count)
 281 {
 282         int i = -1, pop = 0;
 283
 284         while ((i = find_next_bit(map, count, i + 1)) < count)
 285                 pop++;
 286         return pop;
 287 }
 288
 289 static void o2hb_write_timeout(struct work_struct *work)
 290 {
 291         int failed, quorum;
 292         unsigned long flags;
 293         struct o2hb_region *reg =
 294                 container_of(work, struct o2hb_region,
 295                              hr_write_timeout_work.work);
 296
 297         mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u "
 298              "milliseconds\n", reg->hr_dev_name,
 299              jiffies_to_msecs(jiffies - reg->hr_last_timeout_start));
 300
 301         if (o2hb_global_heartbeat_active()) {
 302                 spin_lock_irqsave(&o2hb_live_lock, flags);
 303                 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
 304                         set_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
 305                 failed = o2hb_pop_count(&o2hb_failed_region_bitmap,
 306                                         O2NM_MAX_REGIONS);
 307                 quorum = o2hb_pop_count(&o2hb_quorum_region_bitmap,
 308                                         O2NM_MAX_REGIONS);
 309                 spin_unlock_irqrestore(&o2hb_live_lock, flags);
 310
 311                 mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n",
 312                      quorum, failed);
 313
 314                 /*
 315                  * Fence if the number of failed regions >= half the number
 316                  * of  quorum regions
 317                  */
 318                 if ((failed << 1) < quorum)
 319                         return;
 320         }
 321
 322         o2quo_disk_timeout();
 323 }
 324
 325 static void o2hb_arm_write_timeout(struct o2hb_region *reg)
 326 {
 327         mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n",
 328              O2HB_MAX_WRITE_TIMEOUT_MS);
 329
 330         if (o2hb_global_heartbeat_active()) {
 331                 spin_lock(&o2hb_live_lock);
 332                 clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap);
 333                 spin_unlock(&o2hb_live_lock);
 334         }
 335         cancel_delayed_work(&reg->hr_write_timeout_work);
 336         reg->hr_last_timeout_start = jiffies;
 337         schedule_delayed_work(&reg->hr_write_timeout_work,
 338                               msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS));
 339 }
 340
 341 static void o2hb_disarm_write_timeout(struct o2hb_region *reg)
 342 {
 343         cancel_delayed_work_sync(&reg->hr_write_timeout_work);
 344 }
 345
 346 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc)
 347 {
 348         atomic_set(&wc->wc_num_reqs, 1);
 349         init_completion(&wc->wc_io_complete);
 350         wc->wc_error = 0;
 351 }
 352
 353 /* Used in error paths too */
 354 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc,
 355                                      unsigned int num)
 356 {
 357         /* sadly atomic_sub_and_test() isn't available on all platforms.  The
 358          * good news is that the fast path only completes one at a time */
 359         while(num--) {
 360                 if (atomic_dec_and_test(&wc->wc_num_reqs)) {
 361                         BUG_ON(num > 0);
 362                         complete(&wc->wc_io_complete);
 363                 }
 364         }
 365 }
 366
 367 static void o2hb_wait_on_io(struct o2hb_region *reg,
 368                             struct o2hb_bio_wait_ctxt *wc)
 369 {
 370         o2hb_bio_wait_dec(wc, 1);
 371         wait_for_completion(&wc->wc_io_complete);
 372 }
 373
 374 static void o2hb_bio_end_io(struct bio *bio,
 375                            int error)
 376 {
 377         struct o2hb_bio_wait_ctxt *wc = bio->bi_private;
 378
 379         if (error) {
 380                 mlog(ML_ERROR, "IO Error %d\n", error);
 381                 wc->wc_error = error;
 382         }
 383
 384         o2hb_bio_wait_dec(wc, 1);
 385         bio_put(bio);
 386 }
 387
 388 /* Setup a Bio to cover I/O against num_slots slots starting at
 389  * start_slot. */
 390 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg,
 391                                       struct o2hb_bio_wait_ctxt *wc,
 392                                       unsigned int *current_slot,
 393                                       unsigned int max_slots)
 394 {
 395         int len, current_page;
 396         unsigned int vec_len, vec_start;
 397         unsigned int bits = reg->hr_block_bits;
 398         unsigned int spp = reg->hr_slots_per_page;
 399         unsigned int cs = *current_slot;
 400         struct bio *bio;
 401         struct page *page;
 402
 403         /* Testing has shown this allocation to take long enough under
 404          * GFP_KERNEL that the local node can get fenced. It would be
 405          * nicest if we could pre-allocate these bios and avoid this
 406          * all together. */
 407         bio = bio_alloc(GFP_ATOMIC, 16);
 408         if (!bio) {
 409                 mlog(ML_ERROR, "Could not alloc slots BIO!\n");
 410                 bio = ERR_PTR(-ENOMEM);
 411                 goto bail;
 412         }
 413
 414         /* Must put everything in 512 byte sectors for the bio... */
 415         bio->bi_sector = (reg->hr_start_block + cs) << (bits - 9);
 416         bio->bi_bdev = reg->hr_bdev;
 417         bio->bi_private = wc;
 418         bio->bi_end_io = o2hb_bio_end_io;
 419
 420         vec_start = (cs << bits) % PAGE_CACHE_SIZE;
 421         while(cs < max_slots) {
 422                 current_page = cs / spp;
 423                 page = reg->hr_slot_data[current_page];
 424
 425                 vec_len = min(PAGE_CACHE_SIZE - vec_start,
 426                               (max_slots-cs) * (PAGE_CACHE_SIZE/spp) );
 427
 428                 mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n",
 429                      current_page, vec_len, vec_start);
 430
 431                 len = bio_add_page(bio, page, vec_len, vec_start);
 432                 if (len != vec_len) break;
 433
 434                 cs += vec_len / (PAGE_CACHE_SIZE/spp);
 435                 vec_start = 0;
 436         }
 437
 438 bail:
 439         *current_slot = cs;
 440         return bio;
 441 }
 442
 443 static int o2hb_read_slots(struct o2hb_region *reg,
 444                            unsigned int max_slots)
 445 {
 446         unsigned int current_slot=0;
 447         int status;
 448         struct o2hb_bio_wait_ctxt wc;
 449         struct bio *bio;
 450
 451         o2hb_bio_wait_init(&wc);
 452
 453         while(current_slot < max_slots) {
 454                 bio = o2hb_setup_one_bio(reg, &wc, &current_slot, max_slots);
 455                 if (IS_ERR(bio)) {
 456                         status = PTR_ERR(bio);
 457                         mlog_errno(status);
 458                         goto bail_and_wait;
 459                 }
 460
 461                 atomic_inc(&wc.wc_num_reqs);
 462                 submit_bio(READ, bio);
 463         }
 464
 465         status = 0;
 466
 467 bail_and_wait:
 468         o2hb_wait_on_io(reg, &wc);
 469         if (wc.wc_error && !status)
 470                 status = wc.wc_error;
 471
 472         return status;
 473 }
 474
 475 static int o2hb_issue_node_write(struct o2hb_region *reg,
 476                                  struct o2hb_bio_wait_ctxt *write_wc)
 477 {
 478         int status;
 479         unsigned int slot;
 480         struct bio *bio;
 481
 482         o2hb_bio_wait_init(write_wc);
 483
 484         slot = o2nm_this_node();
 485
 486         bio = o2hb_setup_one_bio(reg, write_wc, &slot, slot+1);
 487         if (IS_ERR(bio)) {
 488                 status = PTR_ERR(bio);
 489                 mlog_errno(status);
 490                 goto bail;
 491         }
 492
 493         atomic_inc(&write_wc->wc_num_reqs);
 494         submit_bio(WRITE, bio);
 495
 496         status = 0;
 497 bail:
 498         return status;
 499 }
 500
 501 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg,
 502                                      struct o2hb_disk_heartbeat_block *hb_block)
 503 {
 504         __le32 old_cksum;
 505         u32 ret;
 506
 507         /* We want to compute the block crc with a 0 value in the
 508          * hb_cksum field. Save it off here and replace after the
 509          * crc. */
 510         old_cksum = hb_block->hb_cksum;
 511         hb_block->hb_cksum = 0;
 512
 513         ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes);
 514
 515         hb_block->hb_cksum = old_cksum;
 516
 517         return ret;
 518 }
 519
 520 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block)
 521 {
 522         mlog(ML_ERROR, "Dump slot information: seq = 0x%llx, node = %u, "
 523              "cksum = 0x%x, generation 0x%llx\n",
 524              (long long)le64_to_cpu(hb_block->hb_seq),
 525              hb_block->hb_node, le32_to_cpu(hb_block->hb_cksum),
 526              (long long)le64_to_cpu(hb_block->hb_generation));
 527 }
 528
 529 static int o2hb_verify_crc(struct o2hb_region *reg,
 530                            struct o2hb_disk_heartbeat_block *hb_block)
 531 {
 532         u32 read, computed;
 533
 534         read = le32_to_cpu(hb_block->hb_cksum);
 535         computed = o2hb_compute_block_crc_le(reg, hb_block);
 536
 537         return read == computed;
 538 }
 539
 540 /* We want to make sure that nobody is heartbeating on top of us --
 541  * this will help detect an invalid configuration. */
 542 static void o2hb_check_last_timestamp(struct o2hb_region *reg)
 543 {
 544         struct o2hb_disk_slot *slot;
 545         struct o2hb_disk_heartbeat_block *hb_block;
 546         char *errstr;
 547
 548         slot = &reg->hr_slots[o2nm_this_node()];
 549         /* Don't check on our 1st timestamp */
 550         if (!slot->ds_last_time)
 551                 return;
 552
 553         hb_block = slot->ds_raw_block;
 554         if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time &&
 555             le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation &&
 556             hb_block->hb_node == slot->ds_node_num)
 557                 return;
 558
 559 #define ERRSTR1         "Another node is heartbeating on device"
 560 #define ERRSTR2         "Heartbeat generation mismatch on device"
 561 #define ERRSTR3         "Heartbeat sequence mismatch on device"
 562
 563         if (hb_block->hb_node != slot->ds_node_num)
 564                 errstr = ERRSTR1;
 565         else if (le64_to_cpu(hb_block->hb_generation) !=
 566                  slot->ds_last_generation)
 567                 errstr = ERRSTR2;
 568         else
 569                 errstr = ERRSTR3;
 570
 571         mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), "
 572              "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name,
 573              slot->ds_node_num, (unsigned long long)slot->ds_last_generation,
 574              (unsigned long long)slot->ds_last_time, hb_block->hb_node,
 575              (unsigned long long)le64_to_cpu(hb_block->hb_generation),
 576              (unsigned long long)le64_to_cpu(hb_block->hb_seq));
 577 }
 578
 579 static inline void o2hb_prepare_block(struct o2hb_region *reg,
 580                                       u64 generation)
 581 {
 582         int node_num;
 583         u64 cputime;
 584         struct o2hb_disk_slot *slot;
 585         struct o2hb_disk_heartbeat_block *hb_block;
 586
 587         node_num = o2nm_this_node();
 588         slot = &reg->hr_slots[node_num];
 589
 590         hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block;
 591         memset(hb_block, 0, reg->hr_block_bytes);
 592         /* TODO: time stuff */
 593         cputime = CURRENT_TIME.tv_sec;
 594         if (!cputime)
 595                 cputime = 1;
 596
 597         hb_block->hb_seq = cpu_to_le64(cputime);
 598         hb_block->hb_node = node_num;
 599         hb_block->hb_generation = cpu_to_le64(generation);
 600         hb_block->hb_dead_ms = cpu_to_le32(o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS);
 601
 602         /* This step must always happen last! */
 603         hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg,
 604                                                                    hb_block));
 605
 606         mlog(ML_HB_BIO, "our node generation = 0x%llx, cksum = 0x%x\n",
 607              (long long)generation,
 608              le32_to_cpu(hb_block->hb_cksum));
 609 }
 610
 611 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall,
 612                                 struct o2nm_node *node,
 613                                 int idx)
 614 {
 615         struct list_head *iter;
 616         struct o2hb_callback_func *f;
 617
 618         list_for_each(iter, &hbcall->list) {
 619                 f = list_entry(iter, struct o2hb_callback_func, hc_item);
 620                 mlog(ML_HEARTBEAT, "calling funcs %p\n", f);
 621                 (f->hc_func)(node, idx, f->hc_data);
 622         }
 623 }
 624
 625 /* Will run the list in order until we process the passed event */
 626 static void o2hb_run_event_list(struct o2hb_node_event *queued_event)
 627 {
 628         int empty;
 629         struct o2hb_callback *hbcall;
 630         struct o2hb_node_event *event;
 631
 632         spin_lock(&o2hb_live_lock);
 633         empty = list_empty(&queued_event->hn_item);
 634         spin_unlock(&o2hb_live_lock);
 635         if (empty)
 636                 return;
 637
 638         /* Holding callback sem assures we don't alter the callback
 639          * lists when doing this, and serializes ourselves with other
 640          * processes wanting callbacks. */
 641         down_write(&o2hb_callback_sem);
 642
 643         spin_lock(&o2hb_live_lock);
 644         while (!list_empty(&o2hb_node_events)
 645                && !list_empty(&queued_event->hn_item)) {
 646                 event = list_entry(o2hb_node_events.next,
 647                                    struct o2hb_node_event,
 648                                    hn_item);
 649                 list_del_init(&event->hn_item);
 650                 spin_unlock(&o2hb_live_lock);
 651
 652                 mlog(ML_HEARTBEAT, "Node %s event for %d\n",
 653                      event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN",
 654                      event->hn_node_num);
 655
 656                 hbcall = hbcall_from_type(event->hn_event_type);
 657
 658                 /* We should *never* have gotten on to the list with a
 659                  * bad type... This isn't something that we should try
 660                  * to recover from. */
 661                 BUG_ON(IS_ERR(hbcall));
 662
 663                 o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num);
 664
 665                 spin_lock(&o2hb_live_lock);
 666         }
 667         spin_unlock(&o2hb_live_lock);
 668
 669         up_write(&o2hb_callback_sem);
 670 }
 671
 672 static void o2hb_queue_node_event(struct o2hb_node_event *event,
 673                                   enum o2hb_callback_type type,
 674                                   struct o2nm_node *node,
 675                                   int node_num)
 676 {
 677         assert_spin_locked(&o2hb_live_lock);
 678
 679         BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB));
 680
 681         event->hn_event_type = type;
 682         event->hn_node = node;
 683         event->hn_node_num = node_num;
 684
 685         mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n",
 686              type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num);
 687
 688         list_add_tail(&event->hn_item, &o2hb_node_events);
 689 }
 690
 691 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot)
 692 {
 693         struct o2hb_node_event event =
 694                 { .hn_item = LIST_HEAD_INIT(event.hn_item), };
 695         struct o2nm_node *node;
 696
 697         node = o2nm_get_node_by_num(slot->ds_node_num);
 698         if (!node)
 699                 return;
 700
 701         spin_lock(&o2hb_live_lock);
 702         if (!list_empty(&slot->ds_live_item)) {
 703                 mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n",
 704                      slot->ds_node_num);
 705
 706                 list_del_init(&slot->ds_live_item);
 707
 708                 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
 709                         clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
 710
 711                         o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node,
 712                                               slot->ds_node_num);
 713                 }
 714         }
 715         spin_unlock(&o2hb_live_lock);
 716
 717         o2hb_run_event_list(&event);
 718
 719         o2nm_node_put(node);
 720 }
 721
 722 static void o2hb_set_quorum_device(struct o2hb_region *reg,
 723                                    struct o2hb_disk_slot *slot)
 724 {
 725         assert_spin_locked(&o2hb_live_lock);
 726
 727         if (!o2hb_global_heartbeat_active())
 728                 return;
 729
 730         if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
 731                 return;
 732
 733         /*
 734          * A region can be added to the quorum only when it sees all
 735          * live nodes heartbeat on it. In other words, the region has been
 736          * added to all nodes.
 737          */
 738         if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap,
 739                    sizeof(o2hb_live_node_bitmap)))
 740                 return;
 741
 742         if (slot->ds_changed_samples < O2HB_LIVE_THRESHOLD)
 743                 return;
 744
 745         printk(KERN_NOTICE "o2hb: Region %s is now a quorum device\n",
 746                config_item_name(&reg->hr_item));
 747
 748         set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
 749
 750         /*
 751          * If global heartbeat active, unpin all regions if the
 752          * region count > CUT_OFF
 753          */
 754         if (o2hb_pop_count(&o2hb_quorum_region_bitmap,
 755                            O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF)
 756                 o2hb_region_unpin(NULL);
 757 }
 758
 759 static int o2hb_check_slot(struct o2hb_region *reg,
 760                            struct o2hb_disk_slot *slot)
 761 {
 762         int changed = 0, gen_changed = 0;
 763         struct o2hb_node_event event =
 764                 { .hn_item = LIST_HEAD_INIT(event.hn_item), };
 765         struct o2nm_node *node;
 766         struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block;
 767         u64 cputime;
 768         unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS;
 769         unsigned int slot_dead_ms;
 770         int tmp;
 771
 772         memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes);
 773
 774         /*
 775          * If a node is no longer configured but is still in the livemap, we
 776          * may need to clear that bit from the livemap.
 777          */
 778         node = o2nm_get_node_by_num(slot->ds_node_num);
 779         if (!node) {
 780                 spin_lock(&o2hb_live_lock);
 781                 tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap);
 782                 spin_unlock(&o2hb_live_lock);
 783                 if (!tmp)
 784                         return 0;
 785         }
 786
 787         if (!o2hb_verify_crc(reg, hb_block)) {
 788                 /* all paths from here will drop o2hb_live_lock for
 789                  * us. */
 790                 spin_lock(&o2hb_live_lock);
 791
 792                 /* Don't print an error on the console in this case -
 793                  * a freshly formatted heartbeat area will not have a
 794                  * crc set on it. */
 795                 if (list_empty(&slot->ds_live_item))
 796                         goto out;
 797
 798                 /* The node is live but pushed out a bad crc. We
 799                  * consider it a transient miss but don't populate any
 800                  * other values as they may be junk. */
 801                 mlog(ML_ERROR, "Node %d has written a bad crc to %s\n",
 802                      slot->ds_node_num, reg->hr_dev_name);
 803                 o2hb_dump_slot(hb_block);
 804
 805                 slot->ds_equal_samples++;
 806                 goto fire_callbacks;
 807         }
 808
 809         /* we don't care if these wrap.. the state transitions below
 810          * clear at the right places */
 811         cputime = le64_to_cpu(hb_block->hb_seq);
 812         if (slot->ds_last_time != cputime)
 813                 slot->ds_changed_samples++;
 814         else
 815                 slot->ds_equal_samples++;
 816         slot->ds_last_time = cputime;
 817
 818         /* The node changed heartbeat generations. We assume this to
 819          * mean it dropped off but came back before we timed out. We
 820          * want to consider it down for the time being but don't want
 821          * to lose any changed_samples state we might build up to
 822          * considering it live again. */
 823         if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) {
 824                 gen_changed = 1;
 825                 slot->ds_equal_samples = 0;
 826                 mlog(ML_HEARTBEAT, "Node %d changed generation (0x%llx "
 827                      "to 0x%llx)\n", slot->ds_node_num,
 828                      (long long)slot->ds_last_generation,
 829                      (long long)le64_to_cpu(hb_block->hb_generation));
 830         }
 831
 832         slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
 833
 834         mlog(ML_HEARTBEAT, "Slot %d gen 0x%llx cksum 0x%x "
 835              "seq %llu last %llu changed %u equal %u\n",
 836              slot->ds_node_num, (long long)slot->ds_last_generation,
 837              le32_to_cpu(hb_block->hb_cksum),
 838              (unsigned long long)le64_to_cpu(hb_block->hb_seq),
 839              (unsigned long long)slot->ds_last_time, slot->ds_changed_samples,
 840              slot->ds_equal_samples);
 841
 842         spin_lock(&o2hb_live_lock);
 843
 844 fire_callbacks:
 845         /* dead nodes only come to life after some number of
 846          * changes at any time during their dead time */
 847         if (list_empty(&slot->ds_live_item) &&
 848             slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) {
 849                 mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n",
 850                      slot->ds_node_num, (long long)slot->ds_last_generation);
 851
 852                 set_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
 853
 854                 /* first on the list generates a callback */
 855                 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
 856                         mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes "
 857                              "bitmap\n", slot->ds_node_num);
 858                         set_bit(slot->ds_node_num, o2hb_live_node_bitmap);
 859
 860                         o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node,
 861                                               slot->ds_node_num);
 862
 863                         changed = 1;
 864                 }
 865
 866                 list_add_tail(&slot->ds_live_item,
 867                               &o2hb_live_slots[slot->ds_node_num]);
 868
 869                 slot->ds_equal_samples = 0;
 870
 871                 /* We want to be sure that all nodes agree on the
 872                  * number of milliseconds before a node will be
 873                  * considered dead. The self-fencing timeout is
 874                  * computed from this value, and a discrepancy might
 875                  * result in heartbeat calling a node dead when it
 876                  * hasn't self-fenced yet. */
 877                 slot_dead_ms = le32_to_cpu(hb_block->hb_dead_ms);
 878                 if (slot_dead_ms && slot_dead_ms != dead_ms) {
 879                         /* TODO: Perhaps we can fail the region here. */
 880                         mlog(ML_ERROR, "Node %d on device %s has a dead count "
 881                              "of %u ms, but our count is %u ms.\n"
 882                              "Please double check your configuration values "
 883                              "for 'O2CB_HEARTBEAT_THRESHOLD'\n",
 884                              slot->ds_node_num, reg->hr_dev_name, slot_dead_ms,
 885                              dead_ms);
 886                 }
 887                 goto out;
 888         }
 889
 890         /* if the list is dead, we're done.. */
 891         if (list_empty(&slot->ds_live_item))
 892                 goto out;
 893
 894         /* live nodes only go dead after enough consequtive missed
 895          * samples..  reset the missed counter whenever we see
 896          * activity */
 897         if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) {
 898                 mlog(ML_HEARTBEAT, "Node %d left my region\n",
 899                      slot->ds_node_num);
 900
 901                 clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap);
 902
 903                 /* last off the live_slot generates a callback */
 904                 list_del_init(&slot->ds_live_item);
 905                 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) {
 906                         mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live "
 907                              "nodes bitmap\n", slot->ds_node_num);
 908                         clear_bit(slot->ds_node_num, o2hb_live_node_bitmap);
 909
 910                         /* node can be null */
 911                         o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB,
 912                                               node, slot->ds_node_num);
 913
 914                         changed = 1;
 915                 }
 916
 917                 /* We don't clear this because the node is still
 918                  * actually writing new blocks. */
 919                 if (!gen_changed)
 920                         slot->ds_changed_samples = 0;
 921                 goto out;
 922         }
 923         if (slot->ds_changed_samples) {
 924                 slot->ds_changed_samples = 0;
 925                 slot->ds_equal_samples = 0;
 926         }
 927 out:
 928         o2hb_set_quorum_device(reg, slot);
 929
 930         spin_unlock(&o2hb_live_lock);
 931
 932         o2hb_run_event_list(&event);
 933
 934         if (node)
 935                 o2nm_node_put(node);
 936         return changed;
 937 }
 938
 939 /* This could be faster if we just implmented a find_last_bit, but I
 940  * don't think the circumstances warrant it. */
 941 static int o2hb_highest_node(unsigned long *nodes,
 942                              int numbits)
 943 {
 944         int highest, node;
 945
 946         highest = numbits;
 947         node = -1;
 948         while ((node = find_next_bit(nodes, numbits, node + 1)) != -1) {
 949                 if (node >= numbits)
 950                         break;
 951
 952                 highest = node;
 953         }
 954
 955         return highest;
 956 }
 957
 958 static int o2hb_do_disk_heartbeat(struct o2hb_region *reg)
 959 {
 960         int i, ret, highest_node, change = 0;
 961         unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)];
 962         unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)];
 963         struct o2hb_bio_wait_ctxt write_wc;
 964
 965         ret = o2nm_configured_node_map(configured_nodes,
 966                                        sizeof(configured_nodes));
 967         if (ret) {
 968                 mlog_errno(ret);
 969                 return ret;
 970         }
 971
 972         /*
 973          * If a node is not configured but is in the livemap, we still need
 974          * to read the slot so as to be able to remove it from the livemap.
 975          */
 976         o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap));
 977         i = -1;
 978         while ((i = find_next_bit(live_node_bitmap,
 979                                   O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
 980                 set_bit(i, configured_nodes);
 981         }
 982
 983         highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES);
 984         if (highest_node >= O2NM_MAX_NODES) {
 985                 mlog(ML_NOTICE, "ocfs2_heartbeat: no configured nodes found!\n");
 986                 return -EINVAL;
 987         }
 988
 989         /* No sense in reading the slots of nodes that don't exist
 990          * yet. Of course, if the node definitions have holes in them
 991          * then we're reading an empty slot anyway... Consider this
 992          * best-effort. */
 993         ret = o2hb_read_slots(reg, highest_node + 1);
 994         if (ret < 0) {
 995                 mlog_errno(ret);
 996                 return ret;
 997         }
 998
 999         /* With an up to date view of the slots, we can check that no
1000          * other node has been improperly configured to heartbeat in
1001          * our slot. */
1002         o2hb_check_last_timestamp(reg);
1003
1004         /* fill in the proper info for our next heartbeat */
1005         o2hb_prepare_block(reg, reg->hr_generation);
1006
1007         /* And fire off the write. Note that we don't wait on this I/O
1008          * until later. */
1009         ret = o2hb_issue_node_write(reg, &write_wc);
1010         if (ret < 0) {
1011                 mlog_errno(ret);
1012                 return ret;
1013         }
1014
1015         i = -1;
1016         while((i = find_next_bit(configured_nodes,
1017                                  O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) {
1018                 change |= o2hb_check_slot(reg, &reg->hr_slots[i]);
1019         }
1020
1021         /*
1022          * We have to be sure we've advertised ourselves on disk
1023          * before we can go to steady state.  This ensures that
1024          * people we find in our steady state have seen us.
1025          */
1026         o2hb_wait_on_io(reg, &write_wc);
1027         if (write_wc.wc_error) {
1028                 /* Do not re-arm the write timeout on I/O error - we
1029                  * can't be sure that the new block ever made it to
1030                  * disk */
1031                 mlog(ML_ERROR, "Write error %d on device \"%s\"\n",
1032                      write_wc.wc_error, reg->hr_dev_name);
1033                 return write_wc.wc_error;
1034         }
1035
1036         o2hb_arm_write_timeout(reg);
1037
1038         /* let the person who launched us know when things are steady */
1039         if (!change && (atomic_read(&reg->hr_steady_iterations) != 0)) {
1040                 if (atomic_dec_and_test(&reg->hr_steady_iterations))
1041                         wake_up(&o2hb_steady_queue);
1042         }
1043
1044         return 0;
1045 }
1046
1047 /* Subtract b from a, storing the result in a. a *must* have a larger
1048  * value than b. */
1049 static void o2hb_tv_subtract(struct timeval *a,
1050                              struct timeval *b)
1051 {
1052         /* just return 0 when a is after b */
1053         if (a->tv_sec < b->tv_sec ||
1054             (a->tv_sec == b->tv_sec && a->tv_usec < b->tv_usec)) {
1055                 a->tv_sec = 0;
1056                 a->tv_usec = 0;
1057                 return;
1058         }
1059
1060         a->tv_sec -= b->tv_sec;
1061         a->tv_usec -= b->tv_usec;
1062         while ( a->tv_usec < 0 ) {
1063                 a->tv_sec--;
1064                 a->tv_usec += 1000000;
1065         }
1066 }
1067
1068 static unsigned int o2hb_elapsed_msecs(struct timeval *start,
1069                                        struct timeval *end)
1070 {
1071         struct timeval res = *end;
1072
1073         o2hb_tv_subtract(&res, start);
1074
1075         return res.tv_sec * 1000 + res.tv_usec / 1000;
1076 }
1077
1078 /*
1079  * we ride the region ref that the region dir holds.  before the region
1080  * dir is removed and drops it ref it will wait to tear down this
1081  * thread.
1082  */
1083 static int o2hb_thread(void *data)
1084 {
1085         int i, ret;
1086         struct o2hb_region *reg = data;
1087         struct o2hb_bio_wait_ctxt write_wc;
1088         struct timeval before_hb, after_hb;
1089         unsigned int elapsed_msec;
1090
1091         mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n");
1092
1093         set_user_nice(current, -20);
1094
1095         /* Pin node */
1096         o2nm_depend_this_node();
1097
1098         while (!kthread_should_stop() && !reg->hr_unclean_stop) {
1099                 /* We track the time spent inside
1100                  * o2hb_do_disk_heartbeat so that we avoid more than
1101                  * hr_timeout_ms between disk writes. On busy systems
1102                  * this should result in a heartbeat which is less
1103                  * likely to time itself out. */
1104                 do_gettimeofday(&before_hb);
1105
1106                 i = 0;
1107                 do {
1108                         ret = o2hb_do_disk_heartbeat(reg);
1109                 } while (ret && ++i < 2);
1110
1111                 do_gettimeofday(&after_hb);
1112                 elapsed_msec = o2hb_elapsed_msecs(&before_hb, &after_hb);
1113
1114                 mlog(ML_HEARTBEAT,
1115                      "start = %lu.%lu, end = %lu.%lu, msec = %u\n",
1116                      before_hb.tv_sec, (unsigned long) before_hb.tv_usec,
1117                      after_hb.tv_sec, (unsigned long) after_hb.tv_usec,
1118                      elapsed_msec);
1119
1120                 if (elapsed_msec < reg->hr_timeout_ms) {
1121                         /* the kthread api has blocked signals for us so no
1122                          * need to record the return value. */
1123                         msleep_interruptible(reg->hr_timeout_ms - elapsed_msec);
1124                 }
1125         }
1126
1127         o2hb_disarm_write_timeout(reg);
1128
1129         /* unclean stop is only used in very bad situation */
1130         for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++)
1131                 o2hb_shutdown_slot(&reg->hr_slots[i]);
1132
1133         /* Explicit down notification - avoid forcing the other nodes
1134          * to timeout on this region when we could just as easily
1135          * write a clear generation - thus indicating to them that
1136          * this node has left this region.
1137          *
1138          * XXX: Should we skip this on unclean_stop? */
1139         o2hb_prepare_block(reg, 0);
1140         ret = o2hb_issue_node_write(reg, &write_wc);
1141         if (ret == 0) {
1142                 o2hb_wait_on_io(reg, &write_wc);
1143         } else {
1144                 mlog_errno(ret);
1145         }
1146
1147         /* Unpin node */
1148         o2nm_undepend_this_node();
1149
1150         mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread exiting\n");
1151
1152         return 0;
1153 }
1154
1155 #ifdef CONFIG_DEBUG_FS
1156 static int o2hb_debug_open(struct inode *inode, struct file *file)
1157 {
1158         struct o2hb_debug_buf *db = inode->i_private;
1159         struct o2hb_region *reg;
1160         unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)];
1161         char *buf = NULL;
1162         int i = -1;
1163         int out = 0;
1164
1165         /* max_nodes should be the largest bitmap we pass here */
1166         BUG_ON(sizeof(map) < db->db_size);
1167
1168         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
1169         if (!buf)
1170                 goto bail;
1171
1172         switch (db->db_type) {
1173         case O2HB_DB_TYPE_LIVENODES:
1174         case O2HB_DB_TYPE_LIVEREGIONS:
1175         case O2HB_DB_TYPE_QUORUMREGIONS:
1176         case O2HB_DB_TYPE_FAILEDREGIONS:
1177                 spin_lock(&o2hb_live_lock);
1178                 memcpy(map, db->db_data, db->db_size);
1179                 spin_unlock(&o2hb_live_lock);
1180                 break;
1181
1182         case O2HB_DB_TYPE_REGION_LIVENODES:
1183                 spin_lock(&o2hb_live_lock);
1184                 reg = (struct o2hb_region *)db->db_data;
1185                 memcpy(map, reg->hr_live_node_bitmap, db->db_size);
1186                 spin_unlock(&o2hb_live_lock);
1187                 break;
1188
1189         case O2HB_DB_TYPE_REGION_NUMBER:
1190                 reg = (struct o2hb_region *)db->db_data;
1191                 out += snprintf(buf + out, PAGE_SIZE - out, "%d\n",
1192                                 reg->hr_region_num);
1193                 goto done;
1194
1195         case O2HB_DB_TYPE_REGION_ELAPSED_TIME:
1196                 reg = (struct o2hb_region *)db->db_data;
1197                 out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
1198                                 jiffies_to_msecs(jiffies -
1199                                                  reg->hr_last_timeout_start));
1200                 goto done;
1201
1202         case O2HB_DB_TYPE_REGION_PINNED:
1203                 reg = (struct o2hb_region *)db->db_data;
1204                 out += snprintf(buf + out, PAGE_SIZE - out, "%u\n",
1205                                 !!reg->hr_item_pinned);
1206                 goto done;
1207
1208         default:
1209                 goto done;
1210         }
1211
1212         while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len)
1213                 out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i);
1214         out += snprintf(buf + out, PAGE_SIZE - out, "\n");
1215
1216 done:
1217         i_size_write(inode, out);
1218
1219         file->private_data = buf;
1220
1221         return 0;
1222 bail:
1223         return -ENOMEM;
1224 }
1225
1226 static int o2hb_debug_release(struct inode *inode, struct file *file)
1227 {
1228         kfree(file->private_data);
1229         return 0;
1230 }
1231
1232 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1233                                  size_t nbytes, loff_t *ppos)
1234 {
1235         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
1236                                        i_size_read(file->f_mapping->host));
1237 }
1238 #else
1239 static int o2hb_debug_open(struct inode *inode, struct file *file)
1240 {
1241         return 0;
1242 }
1243 static int o2hb_debug_release(struct inode *inode, struct file *file)
1244 {
1245         return 0;
1246 }
1247 static ssize_t o2hb_debug_read(struct file *file, char __user *buf,
1248                                size_t nbytes, loff_t *ppos)
1249 {
1250         return 0;
1251 }
1252 #endif  /* CONFIG_DEBUG_FS */
1253
1254 static const struct file_operations o2hb_debug_fops = {
1255         .open =         o2hb_debug_open,
1256         .release =      o2hb_debug_release,
1257         .read =         o2hb_debug_read,
1258         .llseek =       generic_file_llseek,
1259 };
1260
1261 void o2hb_exit(void)
1262 {
1263         kfree(o2hb_db_livenodes);
1264         kfree(o2hb_db_liveregions);
1265         kfree(o2hb_db_quorumregions);
1266         kfree(o2hb_db_failedregions);
1267         debugfs_remove(o2hb_debug_failedregions);
1268         debugfs_remove(o2hb_debug_quorumregions);
1269         debugfs_remove(o2hb_debug_liveregions);
1270         debugfs_remove(o2hb_debug_livenodes);
1271         debugfs_remove(o2hb_debug_dir);
1272 }
1273
1274 static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir,
1275                                         struct o2hb_debug_buf **db, int db_len,
1276                                         int type, int size, int len, void *data)
1277 {
1278         *db = kmalloc(db_len, GFP_KERNEL);
1279         if (!*db)
1280                 return NULL;
1281
1282         (*db)->db_type = type;
1283         (*db)->db_size = size;
1284         (*db)->db_len = len;
1285         (*db)->db_data = data;
1286
1287         return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db,
1288                                    &o2hb_debug_fops);
1289 }
1290
1291 static int o2hb_debug_init(void)
1292 {
1293         int ret = -ENOMEM;
1294
1295         o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL);
1296         if (!o2hb_debug_dir) {
1297                 mlog_errno(ret);
1298                 goto bail;
1299         }
1300
1301         o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES,
1302                                                  o2hb_debug_dir,
1303                                                  &o2hb_db_livenodes,
1304                                                  sizeof(*o2hb_db_livenodes),
1305                                                  O2HB_DB_TYPE_LIVENODES,
1306                                                  sizeof(o2hb_live_node_bitmap),
1307                                                  O2NM_MAX_NODES,
1308                                                  o2hb_live_node_bitmap);
1309         if (!o2hb_debug_livenodes) {
1310                 mlog_errno(ret);
1311                 goto bail;
1312         }
1313
1314         o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS,
1315                                                    o2hb_debug_dir,
1316                                                    &o2hb_db_liveregions,
1317                                                    sizeof(*o2hb_db_liveregions),
1318                                                    O2HB_DB_TYPE_LIVEREGIONS,
1319                                                    sizeof(o2hb_live_region_bitmap),
1320                                                    O2NM_MAX_REGIONS,
1321                                                    o2hb_live_region_bitmap);
1322         if (!o2hb_debug_liveregions) {
1323                 mlog_errno(ret);
1324                 goto bail;
1325         }
1326
1327         o2hb_debug_quorumregions =
1328                         o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS,
1329                                           o2hb_debug_dir,
1330                                           &o2hb_db_quorumregions,
1331                                           sizeof(*o2hb_db_quorumregions),
1332                                           O2HB_DB_TYPE_QUORUMREGIONS,
1333                                           sizeof(o2hb_quorum_region_bitmap),
1334                                           O2NM_MAX_REGIONS,
1335                                           o2hb_quorum_region_bitmap);
1336         if (!o2hb_debug_quorumregions) {
1337                 mlog_errno(ret);
1338                 goto bail;
1339         }
1340
1341         o2hb_debug_failedregions =
1342                         o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS,
1343                                           o2hb_debug_dir,
1344                                           &o2hb_db_failedregions,
1345                                           sizeof(*o2hb_db_failedregions),
1346                                           O2HB_DB_TYPE_FAILEDREGIONS,
1347                                           sizeof(o2hb_failed_region_bitmap),
1348                                           O2NM_MAX_REGIONS,
1349                                           o2hb_failed_region_bitmap);
1350         if (!o2hb_debug_failedregions) {
1351                 mlog_errno(ret);
1352                 goto bail;
1353         }
1354
1355         ret = 0;
1356 bail:
1357         if (ret)
1358                 o2hb_exit();
1359
1360         return ret;
1361 }
1362
1363 int o2hb_init(void)
1364 {
1365         int i;
1366
1367         for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++)
1368                 INIT_LIST_HEAD(&o2hb_callbacks[i].list);
1369
1370         for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++)
1371                 INIT_LIST_HEAD(&o2hb_live_slots[i]);
1372
1373         INIT_LIST_HEAD(&o2hb_node_events);
1374
1375         memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap));
1376         memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap));
1377         memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap));
1378         memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap));
1379         memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap));
1380
1381         o2hb_dependent_users = 0;
1382
1383         return o2hb_debug_init();
1384 }
1385
1386 /* if we're already in a callback then we're already serialized by the sem */
1387 static void o2hb_fill_node_map_from_callback(unsigned long *map,
1388                                              unsigned bytes)
1389 {
1390         BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long)));
1391
1392         memcpy(map, &o2hb_live_node_bitmap, bytes);
1393 }
1394
1395 /*
1396  * get a map of all nodes that are heartbeating in any regions
1397  */
1398 void o2hb_fill_node_map(unsigned long *map, unsigned bytes)
1399 {
1400         /* callers want to serialize this map and callbacks so that they
1401          * can trust that they don't miss nodes coming to the party */
1402         down_read(&o2hb_callback_sem);
1403         spin_lock(&o2hb_live_lock);
1404         o2hb_fill_node_map_from_callback(map, bytes);
1405         spin_unlock(&o2hb_live_lock);
1406         up_read(&o2hb_callback_sem);
1407 }
1408 EXPORT_SYMBOL_GPL(o2hb_fill_node_map);
1409
1410 /*
1411  * heartbeat configfs bits.  The heartbeat set is a default set under
1412  * the cluster set in nodemanager.c.
1413  */
1414
1415 static struct o2hb_region *to_o2hb_region(struct config_item *item)
1416 {
1417         return item ? container_of(item, struct o2hb_region, hr_item) : NULL;
1418 }
1419
1420 /* drop_item only drops its ref after killing the thread, nothing should
1421  * be using the region anymore.  this has to clean up any state that
1422  * attributes might have built up. */
1423 static void o2hb_region_release(struct config_item *item)
1424 {
1425         int i;
1426         struct page *page;
1427         struct o2hb_region *reg = to_o2hb_region(item);
1428
1429         if (reg->hr_tmp_block)
1430                 kfree(reg->hr_tmp_block);
1431
1432         if (reg->hr_slot_data) {
1433                 for (i = 0; i < reg->hr_num_pages; i++) {
1434                         page = reg->hr_slot_data[i];
1435                         if (page)
1436                                 __free_page(page);
1437                 }
1438                 kfree(reg->hr_slot_data);
1439         }
1440
1441         if (reg->hr_bdev)
1442                 blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1443
1444         if (reg->hr_slots)
1445                 kfree(reg->hr_slots);
1446
1447         kfree(reg->hr_db_regnum);
1448         kfree(reg->hr_db_livenodes);
1449         debugfs_remove(reg->hr_debug_livenodes);
1450         debugfs_remove(reg->hr_debug_regnum);
1451         debugfs_remove(reg->hr_debug_elapsed_time);
1452         debugfs_remove(reg->hr_debug_pinned);
1453         debugfs_remove(reg->hr_debug_dir);
1454
1455         spin_lock(&o2hb_live_lock);
1456         list_del(&reg->hr_all_item);
1457         spin_unlock(&o2hb_live_lock);
1458
1459         kfree(reg);
1460 }
1461
1462 static int o2hb_read_block_input(struct o2hb_region *reg,
1463                                  const char *page,
1464                                  size_t count,
1465                                  unsigned long *ret_bytes,
1466                                  unsigned int *ret_bits)
1467 {
1468         unsigned long bytes;
1469         char *p = (char *)page;
1470
1471         bytes = simple_strtoul(p, &p, 0);
1472         if (!p || (*p && (*p != '\n')))
1473                 return -EINVAL;
1474
1475         /* Heartbeat and fs min / max block sizes are the same. */
1476         if (bytes > 4096 || bytes < 512)
1477                 return -ERANGE;
1478         if (hweight16(bytes) != 1)
1479                 return -EINVAL;
1480
1481         if (ret_bytes)
1482                 *ret_bytes = bytes;
1483         if (ret_bits)
1484                 *ret_bits = ffs(bytes) - 1;
1485
1486         return 0;
1487 }
1488
1489 static ssize_t o2hb_region_block_bytes_read(struct o2hb_region *reg,
1490                                             char *page)
1491 {
1492         return sprintf(page, "%u\n", reg->hr_block_bytes);
1493 }
1494
1495 static ssize_t o2hb_region_block_bytes_write(struct o2hb_region *reg,
1496                                              const char *page,
1497                                              size_t count)
1498 {
1499         int status;
1500         unsigned long block_bytes;
1501         unsigned int block_bits;
1502
1503         if (reg->hr_bdev)
1504                 return -EINVAL;
1505
1506         status = o2hb_read_block_input(reg, page, count,
1507                                        &block_bytes, &block_bits);
1508         if (status)
1509                 return status;
1510
1511         reg->hr_block_bytes = (unsigned int)block_bytes;
1512         reg->hr_block_bits = block_bits;
1513
1514         return count;
1515 }
1516
1517 static ssize_t o2hb_region_start_block_read(struct o2hb_region *reg,
1518                                             char *page)
1519 {
1520         return sprintf(page, "%llu\n", reg->hr_start_block);
1521 }
1522
1523 static ssize_t o2hb_region_start_block_write(struct o2hb_region *reg,
1524                                              const char *page,
1525                                              size_t count)
1526 {
1527         unsigned long long tmp;
1528         char *p = (char *)page;
1529
1530         if (reg->hr_bdev)
1531                 return -EINVAL;
1532
1533         tmp = simple_strtoull(p, &p, 0);
1534         if (!p || (*p && (*p != '\n')))
1535                 return -EINVAL;
1536
1537         reg->hr_start_block = tmp;
1538
1539         return count;
1540 }
1541
1542 static ssize_t o2hb_region_blocks_read(struct o2hb_region *reg,
1543                                        char *page)
1544 {
1545         return sprintf(page, "%d\n", reg->hr_blocks);
1546 }
1547
1548 static ssize_t o2hb_region_blocks_write(struct o2hb_region *reg,
1549                                         const char *page,
1550                                         size_t count)
1551 {
1552         unsigned long tmp;
1553         char *p = (char *)page;
1554
1555         if (reg->hr_bdev)
1556                 return -EINVAL;
1557
1558         tmp = simple_strtoul(p, &p, 0);
1559         if (!p || (*p && (*p != '\n')))
1560                 return -EINVAL;
1561
1562         if (tmp > O2NM_MAX_NODES || tmp == 0)
1563                 return -ERANGE;
1564
1565         reg->hr_blocks = (unsigned int)tmp;
1566
1567         return count;
1568 }
1569
1570 static ssize_t o2hb_region_dev_read(struct o2hb_region *reg,
1571                                     char *page)
1572 {
1573         unsigned int ret = 0;
1574
1575         if (reg->hr_bdev)
1576                 ret = sprintf(page, "%s\n", reg->hr_dev_name);
1577
1578         return ret;
1579 }
1580
1581 static void o2hb_init_region_params(struct o2hb_region *reg)
1582 {
1583         reg->hr_slots_per_page = PAGE_CACHE_SIZE >> reg->hr_block_bits;
1584         reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS;
1585
1586         mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n",
1587              reg->hr_start_block, reg->hr_blocks);
1588         mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n",
1589              reg->hr_block_bytes, reg->hr_block_bits);
1590         mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms);
1591         mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold);
1592 }
1593
1594 static int o2hb_map_slot_data(struct o2hb_region *reg)
1595 {
1596         int i, j;
1597         unsigned int last_slot;
1598         unsigned int spp = reg->hr_slots_per_page;
1599         struct page *page;
1600         char *raw;
1601         struct o2hb_disk_slot *slot;
1602
1603         reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL);
1604         if (reg->hr_tmp_block == NULL) {
1605                 mlog_errno(-ENOMEM);
1606                 return -ENOMEM;
1607         }
1608
1609         reg->hr_slots = kcalloc(reg->hr_blocks,
1610                                 sizeof(struct o2hb_disk_slot), GFP_KERNEL);
1611         if (reg->hr_slots == NULL) {
1612                 mlog_errno(-ENOMEM);
1613                 return -ENOMEM;
1614         }
1615
1616         for(i = 0; i < reg->hr_blocks; i++) {
1617                 slot = &reg->hr_slots[i];
1618                 slot->ds_node_num = i;
1619                 INIT_LIST_HEAD(&slot->ds_live_item);
1620                 slot->ds_raw_block = NULL;
1621         }
1622
1623         reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp;
1624         mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks "
1625                            "at %u blocks per page\n",
1626              reg->hr_num_pages, reg->hr_blocks, spp);
1627
1628         reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *),
1629                                     GFP_KERNEL);
1630         if (!reg->hr_slot_data) {
1631                 mlog_errno(-ENOMEM);
1632                 return -ENOMEM;
1633         }
1634
1635         for(i = 0; i < reg->hr_num_pages; i++) {
1636                 page = alloc_page(GFP_KERNEL);
1637                 if (!page) {
1638                         mlog_errno(-ENOMEM);
1639                         return -ENOMEM;
1640                 }
1641
1642                 reg->hr_slot_data[i] = page;
1643
1644                 last_slot = i * spp;
1645                 raw = page_address(page);
1646                 for (j = 0;
1647                      (j < spp) && ((j + last_slot) < reg->hr_blocks);
1648                      j++) {
1649                         BUG_ON((j + last_slot) >= reg->hr_blocks);
1650
1651                         slot = &reg->hr_slots[j + last_slot];
1652                         slot->ds_raw_block =
1653                                 (struct o2hb_disk_heartbeat_block *) raw;
1654
1655                         raw += reg->hr_block_bytes;
1656                 }
1657         }
1658
1659         return 0;
1660 }
1661
1662 /* Read in all the slots available and populate the tracking
1663  * structures so that we can start with a baseline idea of what's
1664  * there. */
1665 static int o2hb_populate_slot_data(struct o2hb_region *reg)
1666 {
1667         int ret, i;
1668         struct o2hb_disk_slot *slot;
1669         struct o2hb_disk_heartbeat_block *hb_block;
1670
1671         ret = o2hb_read_slots(reg, reg->hr_blocks);
1672         if (ret) {
1673                 mlog_errno(ret);
1674                 goto out;
1675         }
1676
1677         /* We only want to get an idea of the values initially in each
1678          * slot, so we do no verification - o2hb_check_slot will
1679          * actually determine if each configured slot is valid and
1680          * whether any values have changed. */
1681         for(i = 0; i < reg->hr_blocks; i++) {
1682                 slot = &reg->hr_slots[i];
1683                 hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block;
1684
1685                 /* Only fill the values that o2hb_check_slot uses to
1686                  * determine changing slots */
1687                 slot->ds_last_time = le64_to_cpu(hb_block->hb_seq);
1688                 slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation);
1689         }
1690
1691 out:
1692         return ret;
1693 }
1694
1695 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */
1696 static ssize_t o2hb_region_dev_write(struct o2hb_region *reg,
1697                                      const char *page,
1698                                      size_t count)
1699 {
1700         struct task_struct *hb_task;
1701         long fd;
1702         int sectsize;
1703         char *p = (char *)page;
1704         struct file *filp = NULL;
1705         struct inode *inode = NULL;
1706         ssize_t ret = -EINVAL;
1707         int live_threshold;
1708
1709         if (reg->hr_bdev)
1710                 goto out;
1711
1712         /* We can't heartbeat without having had our node number
1713          * configured yet. */
1714         if (o2nm_this_node() == O2NM_MAX_NODES)
1715                 goto out;
1716
1717         fd = simple_strtol(p, &p, 0);
1718         if (!p || (*p && (*p != '\n')))
1719                 goto out;
1720
1721         if (fd < 0 || fd >= INT_MAX)
1722                 goto out;
1723
1724         filp = fget(fd);
1725         if (filp == NULL)
1726                 goto out;
1727
1728         if (reg->hr_blocks == 0 || reg->hr_start_block == 0 ||
1729             reg->hr_block_bytes == 0)
1730                 goto out;
1731
1732         inode = igrab(filp->f_mapping->host);
1733         if (inode == NULL)
1734                 goto out;
1735
1736         if (!S_ISBLK(inode->i_mode))
1737                 goto out;
1738
1739         reg->hr_bdev = I_BDEV(filp->f_mapping->host);
1740         ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL);
1741         if (ret) {
1742                 reg->hr_bdev = NULL;
1743                 goto out;
1744         }
1745         inode = NULL;
1746
1747         bdevname(reg->hr_bdev, reg->hr_dev_name);
1748
1749         sectsize = bdev_logical_block_size(reg->hr_bdev);
1750         if (sectsize != reg->hr_block_bytes) {
1751                 mlog(ML_ERROR,
1752                      "blocksize %u incorrect for device, expected %d",
1753                      reg->hr_block_bytes, sectsize);
1754                 ret = -EINVAL;
1755                 goto out;
1756         }
1757
1758         o2hb_init_region_params(reg);
1759
1760         /* Generation of zero is invalid */
1761         do {
1762                 get_random_bytes(&reg->hr_generation,
1763                                  sizeof(reg->hr_generation));
1764         } while (reg->hr_generation == 0);
1765
1766         ret = o2hb_map_slot_data(reg);
1767         if (ret) {
1768                 mlog_errno(ret);
1769                 goto out;
1770         }
1771
1772         ret = o2hb_populate_slot_data(reg);
1773         if (ret) {
1774                 mlog_errno(ret);
1775                 goto out;
1776         }
1777
1778         INIT_DELAYED_WORK(&reg->hr_write_timeout_work, o2hb_write_timeout);
1779
1780         /*
1781          * A node is considered live after it has beat LIVE_THRESHOLD
1782          * times.  We're not steady until we've given them a chance
1783          * _after_ our first read.
1784          * The default threshold is bare minimum so as to limit the delay
1785          * during mounts. For global heartbeat, the threshold doubled for the
1786          * first region.
1787          */
1788         live_threshold = O2HB_LIVE_THRESHOLD;
1789         if (o2hb_global_heartbeat_active()) {
1790                 spin_lock(&o2hb_live_lock);
1791                 if (o2hb_pop_count(&o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1)
1792                         live_threshold <<= 1;
1793                 spin_unlock(&o2hb_live_lock);
1794         }
1795         atomic_set(&reg->hr_steady_iterations, live_threshold + 1);
1796
1797         hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s",
1798                               reg->hr_item.ci_name);
1799         if (IS_ERR(hb_task)) {
1800                 ret = PTR_ERR(hb_task);
1801                 mlog_errno(ret);
1802                 goto out;
1803         }
1804
1805         spin_lock(&o2hb_live_lock);
1806         reg->hr_task = hb_task;
1807         spin_unlock(&o2hb_live_lock);
1808
1809         ret = wait_event_interruptible(o2hb_steady_queue,
1810                                 atomic_read(&reg->hr_steady_iterations) == 0);
1811         if (ret) {
1812                 /* We got interrupted (hello ptrace!).  Clean up */
1813                 spin_lock(&o2hb_live_lock);
1814                 hb_task = reg->hr_task;
1815                 reg->hr_task = NULL;
1816                 spin_unlock(&o2hb_live_lock);
1817
1818                 if (hb_task)
1819                         kthread_stop(hb_task);
1820                 goto out;
1821         }
1822
1823         /* Ok, we were woken.  Make sure it wasn't by drop_item() */
1824         spin_lock(&o2hb_live_lock);
1825         hb_task = reg->hr_task;
1826         if (o2hb_global_heartbeat_active())
1827                 set_bit(reg->hr_region_num, o2hb_live_region_bitmap);
1828         spin_unlock(&o2hb_live_lock);
1829
1830         if (hb_task)
1831                 ret = count;
1832         else
1833                 ret = -EIO;
1834
1835         if (hb_task && o2hb_global_heartbeat_active())
1836                 printk(KERN_NOTICE "o2hb: Heartbeat started on region %s\n",
1837                        config_item_name(&reg->hr_item));
1838
1839 out:
1840         if (filp)
1841                 fput(filp);
1842         if (inode)
1843                 iput(inode);
1844         if (ret < 0) {
1845                 if (reg->hr_bdev) {
1846                         blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE);
1847                         reg->hr_bdev = NULL;
1848                 }
1849         }
1850         return ret;
1851 }
1852
1853 static ssize_t o2hb_region_pid_read(struct o2hb_region *reg,
1854                                       char *page)
1855 {
1856         pid_t pid = 0;
1857
1858         spin_lock(&o2hb_live_lock);
1859         if (reg->hr_task)
1860                 pid = task_pid_nr(reg->hr_task);
1861         spin_unlock(&o2hb_live_lock);
1862
1863         if (!pid)
1864                 return 0;
1865
1866         return sprintf(page, "%u\n", pid);
1867 }
1868
1869 struct o2hb_region_attribute {
1870         struct configfs_attribute attr;
1871         ssize_t (*show)(struct o2hb_region *, char *);
1872         ssize_t (*store)(struct o2hb_region *, const char *, size_t);
1873 };
1874
1875 static struct o2hb_region_attribute o2hb_region_attr_block_bytes = {
1876         .attr   = { .ca_owner = THIS_MODULE,
1877                     .ca_name = "block_bytes",
1878                     .ca_mode = S_IRUGO | S_IWUSR },
1879         .show   = o2hb_region_block_bytes_read,
1880         .store  = o2hb_region_block_bytes_write,
1881 };
1882
1883 static struct o2hb_region_attribute o2hb_region_attr_start_block = {
1884         .attr   = { .ca_owner = THIS_MODULE,
1885                     .ca_name = "start_block",
1886                     .ca_mode = S_IRUGO | S_IWUSR },
1887         .show   = o2hb_region_start_block_read,
1888         .store  = o2hb_region_start_block_write,
1889 };
1890
1891 static struct o2hb_region_attribute o2hb_region_attr_blocks = {
1892         .attr   = { .ca_owner = THIS_MODULE,
1893                     .ca_name = "blocks",
1894                     .ca_mode = S_IRUGO | S_IWUSR },
1895         .show   = o2hb_region_blocks_read,
1896         .store  = o2hb_region_blocks_write,
1897 };
1898
1899 static struct o2hb_region_attribute o2hb_region_attr_dev = {
1900         .attr   = { .ca_owner = THIS_MODULE,
1901                     .ca_name = "dev",
1902                     .ca_mode = S_IRUGO | S_IWUSR },
1903         .show   = o2hb_region_dev_read,
1904         .store  = o2hb_region_dev_write,
1905 };
1906
1907 static struct o2hb_region_attribute o2hb_region_attr_pid = {
1908        .attr   = { .ca_owner = THIS_MODULE,
1909                    .ca_name = "pid",
1910                    .ca_mode = S_IRUGO | S_IRUSR },
1911        .show   = o2hb_region_pid_read,
1912 };
1913
1914 static struct configfs_attribute *o2hb_region_attrs[] = {
1915         &o2hb_region_attr_block_bytes.attr,
1916         &o2hb_region_attr_start_block.attr,
1917         &o2hb_region_attr_blocks.attr,
1918         &o2hb_region_attr_dev.attr,
1919         &o2hb_region_attr_pid.attr,
1920         NULL,
1921 };
1922
1923 static ssize_t o2hb_region_show(struct config_item *item,
1924                                 struct configfs_attribute *attr,
1925                                 char *page)
1926 {
1927         struct o2hb_region *reg = to_o2hb_region(item);
1928         struct o2hb_region_attribute *o2hb_region_attr =
1929                 container_of(attr, struct o2hb_region_attribute, attr);
1930         ssize_t ret = 0;
1931
1932         if (o2hb_region_attr->show)
1933                 ret = o2hb_region_attr->show(reg, page);
1934         return ret;
1935 }
1936
1937 static ssize_t o2hb_region_store(struct config_item *item,
1938                                  struct configfs_attribute *attr,
1939                                  const char *page, size_t count)
1940 {
1941         struct o2hb_region *reg = to_o2hb_region(item);
1942         struct o2hb_region_attribute *o2hb_region_attr =
1943                 container_of(attr, struct o2hb_region_attribute, attr);
1944         ssize_t ret = -EINVAL;
1945
1946         if (o2hb_region_attr->store)
1947                 ret = o2hb_region_attr->store(reg, page, count);
1948         return ret;
1949 }
1950
1951 static struct configfs_item_operations o2hb_region_item_ops = {
1952         .release                = o2hb_region_release,
1953         .show_attribute         = o2hb_region_show,
1954         .store_attribute        = o2hb_region_store,
1955 };
1956
1957 static struct config_item_type o2hb_region_type = {
1958         .ct_item_ops    = &o2hb_region_item_ops,
1959         .ct_attrs       = o2hb_region_attrs,
1960         .ct_owner       = THIS_MODULE,
1961 };
1962
1963 /* heartbeat set */
1964
1965 struct o2hb_heartbeat_group {
1966         struct config_group hs_group;
1967         /* some stuff? */
1968 };
1969
1970 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group)
1971 {
1972         return group ?
1973                 container_of(group, struct o2hb_heartbeat_group, hs_group)
1974                 : NULL;
1975 }
1976
1977 static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir)
1978 {
1979         int ret = -ENOMEM;
1980
1981         reg->hr_debug_dir =
1982                 debugfs_create_dir(config_item_name(&reg->hr_item), dir);
1983         if (!reg->hr_debug_dir) {
1984                 mlog_errno(ret);
1985                 goto bail;
1986         }
1987
1988         reg->hr_debug_livenodes =
1989                         o2hb_debug_create(O2HB_DEBUG_LIVENODES,
1990                                           reg->hr_debug_dir,
1991                                           &(reg->hr_db_livenodes),
1992                                           sizeof(*(reg->hr_db_livenodes)),
1993                                           O2HB_DB_TYPE_REGION_LIVENODES,
1994                                           sizeof(reg->hr_live_node_bitmap),
1995                                           O2NM_MAX_NODES, reg);
1996         if (!reg->hr_debug_livenodes) {
1997                 mlog_errno(ret);
1998                 goto bail;
1999         }
2000
2001         reg->hr_debug_regnum =
2002                         o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER,
2003                                           reg->hr_debug_dir,
2004                                           &(reg->hr_db_regnum),
2005                                           sizeof(*(reg->hr_db_regnum)),
2006                                           O2HB_DB_TYPE_REGION_NUMBER,
2007                                           0, O2NM_MAX_NODES, reg);
2008         if (!reg->hr_debug_regnum) {
2009                 mlog_errno(ret);
2010                 goto bail;
2011         }
2012
2013         reg->hr_debug_elapsed_time =
2014                         o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME,
2015                                           reg->hr_debug_dir,
2016                                           &(reg->hr_db_elapsed_time),
2017                                           sizeof(*(reg->hr_db_elapsed_time)),
2018                                           O2HB_DB_TYPE_REGION_ELAPSED_TIME,
2019                                           0, 0, reg);
2020         if (!reg->hr_debug_elapsed_time) {
2021                 mlog_errno(ret);
2022                 goto bail;
2023         }
2024
2025         reg->hr_debug_pinned =
2026                         o2hb_debug_create(O2HB_DEBUG_REGION_PINNED,
2027                                           reg->hr_debug_dir,
2028                                           &(reg->hr_db_pinned),
2029                                           sizeof(*(reg->hr_db_pinned)),
2030                                           O2HB_DB_TYPE_REGION_PINNED,
2031                                           0, 0, reg);
2032         if (!reg->hr_debug_pinned) {
2033                 mlog_errno(ret);
2034                 goto bail;
2035         }
2036
2037         ret = 0;
2038 bail:
2039         return ret;
2040 }
2041
2042 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group,
2043                                                           const char *name)
2044 {
2045         struct o2hb_region *reg = NULL;
2046         int ret;
2047
2048         reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL);
2049         if (reg == NULL)
2050                 return ERR_PTR(-ENOMEM);
2051
2052         if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) {
2053                 ret = -ENAMETOOLONG;
2054                 goto free;
2055         }
2056
2057         spin_lock(&o2hb_live_lock);
2058         reg->hr_region_num = 0;
2059         if (o2hb_global_heartbeat_active()) {
2060                 reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap,
2061                                                          O2NM_MAX_REGIONS);
2062                 if (reg->hr_region_num >= O2NM_MAX_REGIONS) {
2063                         spin_unlock(&o2hb_live_lock);
2064                         ret = -EFBIG;
2065                         goto free;
2066                 }
2067                 set_bit(reg->hr_region_num, o2hb_region_bitmap);
2068         }
2069         list_add_tail(&reg->hr_all_item, &o2hb_all_regions);
2070         spin_unlock(&o2hb_live_lock);
2071
2072         config_item_init_type_name(&reg->hr_item, name, &o2hb_region_type);
2073
2074         ret = o2hb_debug_region_init(reg, o2hb_debug_dir);
2075         if (ret) {
2076                 config_item_put(&reg->hr_item);
2077                 goto free;
2078         }
2079
2080         return &reg->hr_item;
2081 free:
2082         kfree(reg);
2083         return ERR_PTR(ret);
2084 }
2085
2086 static void o2hb_heartbeat_group_drop_item(struct config_group *group,
2087                                            struct config_item *item)
2088 {
2089         struct task_struct *hb_task;
2090         struct o2hb_region *reg = to_o2hb_region(item);
2091         int quorum_region = 0;
2092
2093         /* stop the thread when the user removes the region dir */
2094         spin_lock(&o2hb_live_lock);
2095         if (o2hb_global_heartbeat_active()) {
2096                 clear_bit(reg->hr_region_num, o2hb_region_bitmap);
2097                 clear_bit(reg->hr_region_num, o2hb_live_region_bitmap);
2098                 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap))
2099                         quorum_region = 1;
2100                 clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap);
2101         }
2102         hb_task = reg->hr_task;
2103         reg->hr_task = NULL;
2104         reg->hr_item_dropped = 1;
2105         spin_unlock(&o2hb_live_lock);
2106
2107         if (hb_task)
2108                 kthread_stop(hb_task);
2109
2110         /*
2111          * If we're racing a dev_write(), we need to wake them.  They will
2112          * check reg->hr_task
2113          */
2114         if (atomic_read(&reg->hr_steady_iterations) != 0) {
2115                 atomic_set(&reg->hr_steady_iterations, 0);
2116                 wake_up(&o2hb_steady_queue);
2117         }
2118
2119         if (o2hb_global_heartbeat_active())
2120                 printk(KERN_NOTICE "o2hb: Heartbeat stopped on region %s\n",
2121                        config_item_name(&reg->hr_item));
2122
2123         config_item_put(item);
2124
2125         if (!o2hb_global_heartbeat_active() || !quorum_region)
2126                 return;
2127
2128         /*
2129          * If global heartbeat active and there are dependent users,
2130          * pin all regions if quorum region count <= CUT_OFF
2131          */
2132         spin_lock(&o2hb_live_lock);
2133
2134         if (!o2hb_dependent_users)
2135                 goto unlock;
2136
2137         if (o2hb_pop_count(&o2hb_quorum_region_bitmap,
2138                            O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2139                 o2hb_region_pin(NULL);
2140
2141 unlock:
2142         spin_unlock(&o2hb_live_lock);
2143 }
2144
2145 struct o2hb_heartbeat_group_attribute {
2146         struct configfs_attribute attr;
2147         ssize_t (*show)(struct o2hb_heartbeat_group *, char *);
2148         ssize_t (*store)(struct o2hb_heartbeat_group *, const char *, size_t);
2149 };
2150
2151 static ssize_t o2hb_heartbeat_group_show(struct config_item *item,
2152                                          struct configfs_attribute *attr,
2153                                          char *page)
2154 {
2155         struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
2156         struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
2157                 container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
2158         ssize_t ret = 0;
2159
2160         if (o2hb_heartbeat_group_attr->show)
2161                 ret = o2hb_heartbeat_group_attr->show(reg, page);
2162         return ret;
2163 }
2164
2165 static ssize_t o2hb_heartbeat_group_store(struct config_item *item,
2166                                           struct configfs_attribute *attr,
2167                                           const char *page, size_t count)
2168 {
2169         struct o2hb_heartbeat_group *reg = to_o2hb_heartbeat_group(to_config_group(item));
2170         struct o2hb_heartbeat_group_attribute *o2hb_heartbeat_group_attr =
2171                 container_of(attr, struct o2hb_heartbeat_group_attribute, attr);
2172         ssize_t ret = -EINVAL;
2173
2174         if (o2hb_heartbeat_group_attr->store)
2175                 ret = o2hb_heartbeat_group_attr->store(reg, page, count);
2176         return ret;
2177 }
2178
2179 static ssize_t o2hb_heartbeat_group_threshold_show(struct o2hb_heartbeat_group *group,
2180                                                      char *page)
2181 {
2182         return sprintf(page, "%u\n", o2hb_dead_threshold);
2183 }
2184
2185 static ssize_t o2hb_heartbeat_group_threshold_store(struct o2hb_heartbeat_group *group,
2186                                                     const char *page,
2187                                                     size_t count)
2188 {
2189         unsigned long tmp;
2190         char *p = (char *)page;
2191
2192         tmp = simple_strtoul(p, &p, 10);
2193         if (!p || (*p && (*p != '\n')))
2194                 return -EINVAL;
2195
2196         /* this will validate ranges for us. */
2197         o2hb_dead_threshold_set((unsigned int) tmp);
2198
2199         return count;
2200 }
2201
2202 static
2203 ssize_t o2hb_heartbeat_group_mode_show(struct o2hb_heartbeat_group *group,
2204                                        char *page)
2205 {
2206         return sprintf(page, "%s\n",
2207                        o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]);
2208 }
2209
2210 static
2211 ssize_t o2hb_heartbeat_group_mode_store(struct o2hb_heartbeat_group *group,
2212                                         const char *page, size_t count)
2213 {
2214         unsigned int i;
2215         int ret;
2216         size_t len;
2217
2218         len = (page[count - 1] == '\n') ? count - 1 : count;
2219         if (!len)
2220                 return -EINVAL;
2221
2222         for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) {
2223                 if (strnicmp(page, o2hb_heartbeat_mode_desc[i], len))
2224                         continue;
2225
2226                 ret = o2hb_global_hearbeat_mode_set(i);
2227                 if (!ret)
2228                         printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n",
2229                                o2hb_heartbeat_mode_desc[i]);
2230                 return count;
2231         }
2232
2233         return -EINVAL;
2234
2235 }
2236
2237 static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_threshold = {
2238         .attr   = { .ca_owner = THIS_MODULE,
2239                     .ca_name = "dead_threshold",
2240                     .ca_mode = S_IRUGO | S_IWUSR },
2241         .show   = o2hb_heartbeat_group_threshold_show,
2242         .store  = o2hb_heartbeat_group_threshold_store,
2243 };
2244
2245 static struct o2hb_heartbeat_group_attribute o2hb_heartbeat_group_attr_mode = {
2246         .attr   = { .ca_owner = THIS_MODULE,
2247                 .ca_name = "mode",
2248                 .ca_mode = S_IRUGO | S_IWUSR },
2249         .show   = o2hb_heartbeat_group_mode_show,
2250         .store  = o2hb_heartbeat_group_mode_store,
2251 };
2252
2253 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = {
2254         &o2hb_heartbeat_group_attr_threshold.attr,
2255         &o2hb_heartbeat_group_attr_mode.attr,
2256         NULL,
2257 };
2258
2259 static struct configfs_item_operations o2hb_hearbeat_group_item_ops = {
2260         .show_attribute         = o2hb_heartbeat_group_show,
2261         .store_attribute        = o2hb_heartbeat_group_store,
2262 };
2263
2264 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = {
2265         .make_item      = o2hb_heartbeat_group_make_item,
2266         .drop_item      = o2hb_heartbeat_group_drop_item,
2267 };
2268
2269 static struct config_item_type o2hb_heartbeat_group_type = {
2270         .ct_group_ops   = &o2hb_heartbeat_group_group_ops,
2271         .ct_item_ops    = &o2hb_hearbeat_group_item_ops,
2272         .ct_attrs       = o2hb_heartbeat_group_attrs,
2273         .ct_owner       = THIS_MODULE,
2274 };
2275
2276 /* this is just here to avoid touching group in heartbeat.h which the
2277  * entire damn world #includes */
2278 struct config_group *o2hb_alloc_hb_set(void)
2279 {
2280         struct o2hb_heartbeat_group *hs = NULL;
2281         struct config_group *ret = NULL;
2282
2283         hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL);
2284         if (hs == NULL)
2285                 goto out;
2286
2287         config_group_init_type_name(&hs->hs_group, "heartbeat",
2288                                     &o2hb_heartbeat_group_type);
2289
2290         ret = &hs->hs_group;
2291 out:
2292         if (ret == NULL)
2293                 kfree(hs);
2294         return ret;
2295 }
2296
2297 void o2hb_free_hb_set(struct config_group *group)
2298 {
2299         struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group);
2300         kfree(hs);
2301 }
2302
2303 /* hb callback registration and issuing */
2304
2305 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type)
2306 {
2307         if (type == O2HB_NUM_CB)
2308                 return ERR_PTR(-EINVAL);
2309
2310         return &o2hb_callbacks[type];
2311 }
2312
2313 void o2hb_setup_callback(struct o2hb_callback_func *hc,
2314                          enum o2hb_callback_type type,
2315                          o2hb_cb_func *func,
2316                          void *data,
2317                          int priority)
2318 {
2319         INIT_LIST_HEAD(&hc->hc_item);
2320         hc->hc_func = func;
2321         hc->hc_data = data;
2322         hc->hc_priority = priority;
2323         hc->hc_type = type;
2324         hc->hc_magic = O2HB_CB_MAGIC;
2325 }
2326 EXPORT_SYMBOL_GPL(o2hb_setup_callback);
2327
2328 /*
2329  * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2330  * In global heartbeat mode, region_uuid passed is NULL.
2331  *
2332  * In local, we only pin the matching region. In global we pin all the active
2333  * regions.
2334  */
2335 static int o2hb_region_pin(const char *region_uuid)
2336 {
2337         int ret = 0, found = 0;
2338         struct o2hb_region *reg;
2339         char *uuid;
2340
2341         assert_spin_locked(&o2hb_live_lock);
2342
2343         list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2344                 uuid = config_item_name(&reg->hr_item);
2345
2346                 /* local heartbeat */
2347                 if (region_uuid) {
2348                         if (strcmp(region_uuid, uuid))
2349                                 continue;
2350                         found = 1;
2351                 }
2352
2353                 if (reg->hr_item_pinned || reg->hr_item_dropped)
2354                         goto skip_pin;
2355
2356                 /* Ignore ENOENT only for local hb (userdlm domain) */
2357                 ret = o2nm_depend_item(&reg->hr_item);
2358                 if (!ret) {
2359                         mlog(ML_CLUSTER, "Pin region %s\n", uuid);
2360                         reg->hr_item_pinned = 1;
2361                 } else {
2362                         if (ret == -ENOENT && found)
2363                                 ret = 0;
2364                         else {
2365                                 mlog(ML_ERROR, "Pin region %s fails with %d\n",
2366                                      uuid, ret);
2367                                 break;
2368                         }
2369                 }
2370 skip_pin:
2371                 if (found)
2372                         break;
2373         }
2374
2375         return ret;
2376 }
2377
2378 /*
2379  * In local heartbeat mode, region_uuid passed matches the dlm domain name.
2380  * In global heartbeat mode, region_uuid passed is NULL.
2381  *
2382  * In local, we only unpin the matching region. In global we unpin all the
2383  * active regions.
2384  */
2385 static void o2hb_region_unpin(const char *region_uuid)
2386 {
2387         struct o2hb_region *reg;
2388         char *uuid;
2389         int found = 0;
2390
2391         assert_spin_locked(&o2hb_live_lock);
2392
2393         list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2394                 uuid = config_item_name(&reg->hr_item);
2395                 if (region_uuid) {
2396                         if (strcmp(region_uuid, uuid))
2397                                 continue;
2398                         found = 1;
2399                 }
2400
2401                 if (reg->hr_item_pinned) {
2402                         mlog(ML_CLUSTER, "Unpin region %s\n", uuid);
2403                         o2nm_undepend_item(&reg->hr_item);
2404                         reg->hr_item_pinned = 0;
2405                 }
2406                 if (found)
2407                         break;
2408         }
2409 }
2410
2411 static int o2hb_region_inc_user(const char *region_uuid)
2412 {
2413         int ret = 0;
2414
2415         spin_lock(&o2hb_live_lock);
2416
2417         /* local heartbeat */
2418         if (!o2hb_global_heartbeat_active()) {
2419             ret = o2hb_region_pin(region_uuid);
2420             goto unlock;
2421         }
2422
2423         /*
2424          * if global heartbeat active and this is the first dependent user,
2425          * pin all regions if quorum region count <= CUT_OFF
2426          */
2427         o2hb_dependent_users++;
2428         if (o2hb_dependent_users > 1)
2429                 goto unlock;
2430
2431         if (o2hb_pop_count(&o2hb_quorum_region_bitmap,
2432                            O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF)
2433                 ret = o2hb_region_pin(NULL);
2434
2435 unlock:
2436         spin_unlock(&o2hb_live_lock);
2437         return ret;
2438 }
2439
2440 void o2hb_region_dec_user(const char *region_uuid)
2441 {
2442         spin_lock(&o2hb_live_lock);
2443
2444         /* local heartbeat */
2445         if (!o2hb_global_heartbeat_active()) {
2446             o2hb_region_unpin(region_uuid);
2447             goto unlock;
2448         }
2449
2450         /*
2451          * if global heartbeat active and there are no dependent users,
2452          * unpin all quorum regions
2453          */
2454         o2hb_dependent_users--;
2455         if (!o2hb_dependent_users)
2456                 o2hb_region_unpin(NULL);
2457
2458 unlock:
2459         spin_unlock(&o2hb_live_lock);
2460 }
2461
2462 int o2hb_register_callback(const char *region_uuid,
2463                            struct o2hb_callback_func *hc)
2464 {
2465         struct o2hb_callback_func *tmp;
2466         struct list_head *iter;
2467         struct o2hb_callback *hbcall;
2468         int ret;
2469
2470         BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2471         BUG_ON(!list_empty(&hc->hc_item));
2472
2473         hbcall = hbcall_from_type(hc->hc_type);
2474         if (IS_ERR(hbcall)) {
2475                 ret = PTR_ERR(hbcall);
2476                 goto out;
2477         }
2478
2479         if (region_uuid) {
2480                 ret = o2hb_region_inc_user(region_uuid);
2481                 if (ret) {
2482                         mlog_errno(ret);
2483                         goto out;
2484                 }
2485         }
2486
2487         down_write(&o2hb_callback_sem);
2488
2489         list_for_each(iter, &hbcall->list) {
2490                 tmp = list_entry(iter, struct o2hb_callback_func, hc_item);
2491                 if (hc->hc_priority < tmp->hc_priority) {
2492                         list_add_tail(&hc->hc_item, iter);
2493                         break;
2494                 }
2495         }
2496         if (list_empty(&hc->hc_item))
2497                 list_add_tail(&hc->hc_item, &hbcall->list);
2498
2499         up_write(&o2hb_callback_sem);
2500         ret = 0;
2501 out:
2502         mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n",
2503              ret, __builtin_return_address(0), hc);
2504         return ret;
2505 }
2506 EXPORT_SYMBOL_GPL(o2hb_register_callback);
2507
2508 void o2hb_unregister_callback(const char *region_uuid,
2509                               struct o2hb_callback_func *hc)
2510 {
2511         BUG_ON(hc->hc_magic != O2HB_CB_MAGIC);
2512
2513         mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n",
2514              __builtin_return_address(0), hc);
2515
2516         /* XXX Can this happen _with_ a region reference? */
2517         if (list_empty(&hc->hc_item))
2518                 return;
2519
2520         if (region_uuid)
2521                 o2hb_region_dec_user(region_uuid);
2522
2523         down_write(&o2hb_callback_sem);
2524
2525         list_del_init(&hc->hc_item);
2526
2527         up_write(&o2hb_callback_sem);
2528 }
2529 EXPORT_SYMBOL_GPL(o2hb_unregister_callback);
2530
2531 int o2hb_check_node_heartbeating(u8 node_num)
2532 {
2533         unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2534
2535         o2hb_fill_node_map(testing_map, sizeof(testing_map));
2536         if (!test_bit(node_num, testing_map)) {
2537                 mlog(ML_HEARTBEAT,
2538                      "node (%u) does not have heartbeating enabled.\n",
2539                      node_num);
2540                 return 0;
2541         }
2542
2543         return 1;
2544 }
2545 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating);
2546
2547 int o2hb_check_node_heartbeating_from_callback(u8 node_num)
2548 {
2549         unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)];
2550
2551         o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map));
2552         if (!test_bit(node_num, testing_map)) {
2553                 mlog(ML_HEARTBEAT,
2554                      "node (%u) does not have heartbeating enabled.\n",
2555                      node_num);
2556                 return 0;
2557         }
2558
2559         return 1;
2560 }
2561 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback);
2562
2563 /* Makes sure our local node is configured with a node number, and is
2564  * heartbeating. */
2565 int o2hb_check_local_node_heartbeating(void)
2566 {
2567         u8 node_num;
2568
2569         /* if this node was set then we have networking */
2570         node_num = o2nm_this_node();
2571         if (node_num == O2NM_MAX_NODES) {
2572                 mlog(ML_HEARTBEAT, "this node has not been configured.\n");
2573                 return 0;
2574         }
2575
2576         return o2hb_check_node_heartbeating(node_num);
2577 }
2578 EXPORT_SYMBOL_GPL(o2hb_check_local_node_heartbeating);
2579
2580 /*
2581  * this is just a hack until we get the plumbing which flips file systems
2582  * read only and drops the hb ref instead of killing the node dead.
2583  */
2584 void o2hb_stop_all_regions(void)
2585 {
2586         struct o2hb_region *reg;
2587
2588         mlog(ML_ERROR, "stopping heartbeat on all active regions.\n");
2589
2590         spin_lock(&o2hb_live_lock);
2591
2592         list_for_each_entry(reg, &o2hb_all_regions, hr_all_item)
2593                 reg->hr_unclean_stop = 1;
2594
2595         spin_unlock(&o2hb_live_lock);
2596 }
2597 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions);
2598
2599 int o2hb_get_all_regions(char *region_uuids, u8 max_regions)
2600 {
2601         struct o2hb_region *reg;
2602         int numregs = 0;
2603         char *p;
2604
2605         spin_lock(&o2hb_live_lock);
2606
2607         p = region_uuids;
2608         list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) {
2609                 mlog(0, "Region: %s\n", config_item_name(&reg->hr_item));
2610                 if (numregs < max_regions) {
2611                         memcpy(p, config_item_name(&reg->hr_item),
2612                                O2HB_MAX_REGION_NAME_LEN);
2613                         p += O2HB_MAX_REGION_NAME_LEN;
2614                 }
2615                 numregs++;
2616         }
2617
2618         spin_unlock(&o2hb_live_lock);
2619
2620         return numregs;
2621 }
2622 EXPORT_SYMBOL_GPL(o2hb_get_all_regions);
2623
2624 int o2hb_global_heartbeat_active(void)
2625 {
2626         return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL);
2627 }
2628 EXPORT_SYMBOL(o2hb_global_heartbeat_active);