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 #include <linux/bitmap.h> 39 #include <linux/ktime.h> 40 #include "heartbeat.h" 41 #include "tcp.h" 42 #include "nodemanager.h" 43 #include "quorum.h" 44 45 #include "masklog.h" 46 47 48 /* 49 * The first heartbeat pass had one global thread that would serialize all hb 50 * callback calls. This global serializing sem should only be removed once 51 * we've made sure that all callees can deal with being called concurrently 52 * from multiple hb region threads. 53 */ 54 static DECLARE_RWSEM(o2hb_callback_sem); 55 56 /* 57 * multiple hb threads are watching multiple regions. A node is live 58 * whenever any of the threads sees activity from the node in its region. 59 */ 60 static DEFINE_SPINLOCK(o2hb_live_lock); 61 static struct list_head o2hb_live_slots[O2NM_MAX_NODES]; 62 static unsigned long o2hb_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; 63 static LIST_HEAD(o2hb_node_events); 64 static DECLARE_WAIT_QUEUE_HEAD(o2hb_steady_queue); 65 66 /* 67 * In global heartbeat, we maintain a series of region bitmaps. 68 * - o2hb_region_bitmap allows us to limit the region number to max region. 69 * - o2hb_live_region_bitmap tracks live regions (seen steady iterations). 70 * - o2hb_quorum_region_bitmap tracks live regions that have seen all nodes 71 * heartbeat on it. 72 * - o2hb_failed_region_bitmap tracks the regions that have seen io timeouts. 73 */ 74 static unsigned long o2hb_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; 75 static unsigned long o2hb_live_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; 76 static unsigned long o2hb_quorum_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; 77 static unsigned long o2hb_failed_region_bitmap[BITS_TO_LONGS(O2NM_MAX_REGIONS)]; 78 79 #define O2HB_DB_TYPE_LIVENODES 0 80 #define O2HB_DB_TYPE_LIVEREGIONS 1 81 #define O2HB_DB_TYPE_QUORUMREGIONS 2 82 #define O2HB_DB_TYPE_FAILEDREGIONS 3 83 #define O2HB_DB_TYPE_REGION_LIVENODES 4 84 #define O2HB_DB_TYPE_REGION_NUMBER 5 85 #define O2HB_DB_TYPE_REGION_ELAPSED_TIME 6 86 #define O2HB_DB_TYPE_REGION_PINNED 7 87 struct o2hb_debug_buf { 88 int db_type; 89 int db_size; 90 int db_len; 91 void *db_data; 92 }; 93 94 static struct o2hb_debug_buf *o2hb_db_livenodes; 95 static struct o2hb_debug_buf *o2hb_db_liveregions; 96 static struct o2hb_debug_buf *o2hb_db_quorumregions; 97 static struct o2hb_debug_buf *o2hb_db_failedregions; 98 99 #define O2HB_DEBUG_DIR "o2hb" 100 #define O2HB_DEBUG_LIVENODES "livenodes" 101 #define O2HB_DEBUG_LIVEREGIONS "live_regions" 102 #define O2HB_DEBUG_QUORUMREGIONS "quorum_regions" 103 #define O2HB_DEBUG_FAILEDREGIONS "failed_regions" 104 #define O2HB_DEBUG_REGION_NUMBER "num" 105 #define O2HB_DEBUG_REGION_ELAPSED_TIME "elapsed_time_in_ms" 106 #define O2HB_DEBUG_REGION_PINNED "pinned" 107 108 static struct dentry *o2hb_debug_dir; 109 static struct dentry *o2hb_debug_livenodes; 110 static struct dentry *o2hb_debug_liveregions; 111 static struct dentry *o2hb_debug_quorumregions; 112 static struct dentry *o2hb_debug_failedregions; 113 114 static LIST_HEAD(o2hb_all_regions); 115 116 static struct o2hb_callback { 117 struct list_head list; 118 } o2hb_callbacks[O2HB_NUM_CB]; 119 120 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type); 121 122 #define O2HB_DEFAULT_BLOCK_BITS 9 123 124 enum o2hb_heartbeat_modes { 125 O2HB_HEARTBEAT_LOCAL = 0, 126 O2HB_HEARTBEAT_GLOBAL, 127 O2HB_HEARTBEAT_NUM_MODES, 128 }; 129 130 char *o2hb_heartbeat_mode_desc[O2HB_HEARTBEAT_NUM_MODES] = { 131 "local", /* O2HB_HEARTBEAT_LOCAL */ 132 "global", /* O2HB_HEARTBEAT_GLOBAL */ 133 }; 134 135 unsigned int o2hb_dead_threshold = O2HB_DEFAULT_DEAD_THRESHOLD; 136 unsigned int o2hb_heartbeat_mode = O2HB_HEARTBEAT_LOCAL; 137 138 /* 139 * o2hb_dependent_users tracks the number of registered callbacks that depend 140 * on heartbeat. o2net and o2dlm are two entities that register this callback. 141 * However only o2dlm depends on the heartbeat. It does not want the heartbeat 142 * to stop while a dlm domain is still active. 143 */ 144 unsigned int o2hb_dependent_users; 145 146 /* 147 * In global heartbeat mode, all regions are pinned if there are one or more 148 * dependent users and the quorum region count is <= O2HB_PIN_CUT_OFF. All 149 * regions are unpinned if the region count exceeds the cut off or the number 150 * of dependent users falls to zero. 151 */ 152 #define O2HB_PIN_CUT_OFF 3 153 154 /* 155 * In local heartbeat mode, we assume the dlm domain name to be the same as 156 * region uuid. This is true for domains created for the file system but not 157 * necessarily true for userdlm domains. This is a known limitation. 158 * 159 * In global heartbeat mode, we pin/unpin all o2hb regions. This solution 160 * works for both file system and userdlm domains. 161 */ 162 static int o2hb_region_pin(const char *region_uuid); 163 static void o2hb_region_unpin(const char *region_uuid); 164 165 /* Only sets a new threshold if there are no active regions. 166 * 167 * No locking or otherwise interesting code is required for reading 168 * o2hb_dead_threshold as it can't change once regions are active and 169 * it's not interesting to anyone until then anyway. */ 170 static void o2hb_dead_threshold_set(unsigned int threshold) 171 { 172 if (threshold > O2HB_MIN_DEAD_THRESHOLD) { 173 spin_lock(&o2hb_live_lock); 174 if (list_empty(&o2hb_all_regions)) 175 o2hb_dead_threshold = threshold; 176 spin_unlock(&o2hb_live_lock); 177 } 178 } 179 180 static int o2hb_global_heartbeat_mode_set(unsigned int hb_mode) 181 { 182 int ret = -1; 183 184 if (hb_mode < O2HB_HEARTBEAT_NUM_MODES) { 185 spin_lock(&o2hb_live_lock); 186 if (list_empty(&o2hb_all_regions)) { 187 o2hb_heartbeat_mode = hb_mode; 188 ret = 0; 189 } 190 spin_unlock(&o2hb_live_lock); 191 } 192 193 return ret; 194 } 195 196 struct o2hb_node_event { 197 struct list_head hn_item; 198 enum o2hb_callback_type hn_event_type; 199 struct o2nm_node *hn_node; 200 int hn_node_num; 201 }; 202 203 struct o2hb_disk_slot { 204 struct o2hb_disk_heartbeat_block *ds_raw_block; 205 u8 ds_node_num; 206 u64 ds_last_time; 207 u64 ds_last_generation; 208 u16 ds_equal_samples; 209 u16 ds_changed_samples; 210 struct list_head ds_live_item; 211 }; 212 213 /* each thread owns a region.. when we're asked to tear down the region 214 * we ask the thread to stop, who cleans up the region */ 215 struct o2hb_region { 216 struct config_item hr_item; 217 218 struct list_head hr_all_item; 219 unsigned hr_unclean_stop:1, 220 hr_aborted_start:1, 221 hr_item_pinned:1, 222 hr_item_dropped:1, 223 hr_node_deleted:1; 224 225 /* protected by the hr_callback_sem */ 226 struct task_struct *hr_task; 227 228 unsigned int hr_blocks; 229 unsigned long long hr_start_block; 230 231 unsigned int hr_block_bits; 232 unsigned int hr_block_bytes; 233 234 unsigned int hr_slots_per_page; 235 unsigned int hr_num_pages; 236 237 struct page **hr_slot_data; 238 struct block_device *hr_bdev; 239 struct o2hb_disk_slot *hr_slots; 240 241 /* live node map of this region */ 242 unsigned long hr_live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; 243 unsigned int hr_region_num; 244 245 struct dentry *hr_debug_dir; 246 struct dentry *hr_debug_livenodes; 247 struct dentry *hr_debug_regnum; 248 struct dentry *hr_debug_elapsed_time; 249 struct dentry *hr_debug_pinned; 250 struct o2hb_debug_buf *hr_db_livenodes; 251 struct o2hb_debug_buf *hr_db_regnum; 252 struct o2hb_debug_buf *hr_db_elapsed_time; 253 struct o2hb_debug_buf *hr_db_pinned; 254 255 /* let the person setting up hb wait for it to return until it 256 * has reached a 'steady' state. This will be fixed when we have 257 * a more complete api that doesn't lead to this sort of fragility. */ 258 atomic_t hr_steady_iterations; 259 260 /* terminate o2hb thread if it does not reach steady state 261 * (hr_steady_iterations == 0) within hr_unsteady_iterations */ 262 atomic_t hr_unsteady_iterations; 263 264 char hr_dev_name[BDEVNAME_SIZE]; 265 266 unsigned int hr_timeout_ms; 267 268 /* randomized as the region goes up and down so that a node 269 * recognizes a node going up and down in one iteration */ 270 u64 hr_generation; 271 272 struct delayed_work hr_write_timeout_work; 273 unsigned long hr_last_timeout_start; 274 275 /* negotiate timer, used to negotiate extending hb timeout. */ 276 struct delayed_work hr_nego_timeout_work; 277 unsigned long hr_nego_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; 278 279 /* Used during o2hb_check_slot to hold a copy of the block 280 * being checked because we temporarily have to zero out the 281 * crc field. */ 282 struct o2hb_disk_heartbeat_block *hr_tmp_block; 283 284 /* Message key for negotiate timeout message. */ 285 unsigned int hr_key; 286 struct list_head hr_handler_list; 287 288 /* last hb status, 0 for success, other value for error. */ 289 int hr_last_hb_status; 290 }; 291 292 struct o2hb_bio_wait_ctxt { 293 atomic_t wc_num_reqs; 294 struct completion wc_io_complete; 295 int wc_error; 296 }; 297 298 #define O2HB_NEGO_TIMEOUT_MS (O2HB_MAX_WRITE_TIMEOUT_MS/2) 299 300 enum { 301 O2HB_NEGO_TIMEOUT_MSG = 1, 302 O2HB_NEGO_APPROVE_MSG = 2, 303 }; 304 305 struct o2hb_nego_msg { 306 u8 node_num; 307 }; 308 309 static void o2hb_write_timeout(struct work_struct *work) 310 { 311 int failed, quorum; 312 struct o2hb_region *reg = 313 container_of(work, struct o2hb_region, 314 hr_write_timeout_work.work); 315 316 mlog(ML_ERROR, "Heartbeat write timeout to device %s after %u " 317 "milliseconds\n", reg->hr_dev_name, 318 jiffies_to_msecs(jiffies - reg->hr_last_timeout_start)); 319 320 if (o2hb_global_heartbeat_active()) { 321 spin_lock(&o2hb_live_lock); 322 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) 323 set_bit(reg->hr_region_num, o2hb_failed_region_bitmap); 324 failed = bitmap_weight(o2hb_failed_region_bitmap, 325 O2NM_MAX_REGIONS); 326 quorum = bitmap_weight(o2hb_quorum_region_bitmap, 327 O2NM_MAX_REGIONS); 328 spin_unlock(&o2hb_live_lock); 329 330 mlog(ML_HEARTBEAT, "Number of regions %d, failed regions %d\n", 331 quorum, failed); 332 333 /* 334 * Fence if the number of failed regions >= half the number 335 * of quorum regions 336 */ 337 if ((failed << 1) < quorum) 338 return; 339 } 340 341 o2quo_disk_timeout(); 342 } 343 344 static void o2hb_arm_timeout(struct o2hb_region *reg) 345 { 346 /* Arm writeout only after thread reaches steady state */ 347 if (atomic_read(®->hr_steady_iterations) != 0) 348 return; 349 350 mlog(ML_HEARTBEAT, "Queue write timeout for %u ms\n", 351 O2HB_MAX_WRITE_TIMEOUT_MS); 352 353 if (o2hb_global_heartbeat_active()) { 354 spin_lock(&o2hb_live_lock); 355 clear_bit(reg->hr_region_num, o2hb_failed_region_bitmap); 356 spin_unlock(&o2hb_live_lock); 357 } 358 cancel_delayed_work(®->hr_write_timeout_work); 359 schedule_delayed_work(®->hr_write_timeout_work, 360 msecs_to_jiffies(O2HB_MAX_WRITE_TIMEOUT_MS)); 361 362 cancel_delayed_work(®->hr_nego_timeout_work); 363 /* negotiate timeout must be less than write timeout. */ 364 schedule_delayed_work(®->hr_nego_timeout_work, 365 msecs_to_jiffies(O2HB_NEGO_TIMEOUT_MS)); 366 memset(reg->hr_nego_node_bitmap, 0, sizeof(reg->hr_nego_node_bitmap)); 367 } 368 369 static void o2hb_disarm_timeout(struct o2hb_region *reg) 370 { 371 cancel_delayed_work_sync(®->hr_write_timeout_work); 372 cancel_delayed_work_sync(®->hr_nego_timeout_work); 373 } 374 375 static int o2hb_send_nego_msg(int key, int type, u8 target) 376 { 377 struct o2hb_nego_msg msg; 378 int status, ret; 379 380 msg.node_num = o2nm_this_node(); 381 again: 382 ret = o2net_send_message(type, key, &msg, sizeof(msg), 383 target, &status); 384 385 if (ret == -EAGAIN || ret == -ENOMEM) { 386 msleep(100); 387 goto again; 388 } 389 390 return ret; 391 } 392 393 static void o2hb_nego_timeout(struct work_struct *work) 394 { 395 unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; 396 int master_node, i, ret; 397 struct o2hb_region *reg; 398 399 reg = container_of(work, struct o2hb_region, hr_nego_timeout_work.work); 400 /* don't negotiate timeout if last hb failed since it is very 401 * possible io failed. Should let write timeout fence self. 402 */ 403 if (reg->hr_last_hb_status) 404 return; 405 406 o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap)); 407 /* lowest node as master node to make negotiate decision. */ 408 master_node = find_next_bit(live_node_bitmap, O2NM_MAX_NODES, 0); 409 410 if (master_node == o2nm_this_node()) { 411 if (!test_bit(master_node, reg->hr_nego_node_bitmap)) { 412 printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s).\n", 413 o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000, 414 config_item_name(®->hr_item), reg->hr_dev_name); 415 set_bit(master_node, reg->hr_nego_node_bitmap); 416 } 417 if (memcmp(reg->hr_nego_node_bitmap, live_node_bitmap, 418 sizeof(reg->hr_nego_node_bitmap))) { 419 /* check negotiate bitmap every second to do timeout 420 * approve decision. 421 */ 422 schedule_delayed_work(®->hr_nego_timeout_work, 423 msecs_to_jiffies(1000)); 424 425 return; 426 } 427 428 printk(KERN_NOTICE "o2hb: all nodes hb write hung, maybe region %s (%s) is down.\n", 429 config_item_name(®->hr_item), reg->hr_dev_name); 430 /* approve negotiate timeout request. */ 431 o2hb_arm_timeout(reg); 432 433 i = -1; 434 while ((i = find_next_bit(live_node_bitmap, 435 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { 436 if (i == master_node) 437 continue; 438 439 mlog(ML_HEARTBEAT, "send NEGO_APPROVE msg to node %d\n", i); 440 ret = o2hb_send_nego_msg(reg->hr_key, 441 O2HB_NEGO_APPROVE_MSG, i); 442 if (ret) 443 mlog(ML_ERROR, "send NEGO_APPROVE msg to node %d fail %d\n", 444 i, ret); 445 } 446 } else { 447 /* negotiate timeout with master node. */ 448 printk(KERN_NOTICE "o2hb: node %d hb write hung for %ds on region %s (%s), negotiate timeout with node %d.\n", 449 o2nm_this_node(), O2HB_NEGO_TIMEOUT_MS/1000, config_item_name(®->hr_item), 450 reg->hr_dev_name, master_node); 451 ret = o2hb_send_nego_msg(reg->hr_key, O2HB_NEGO_TIMEOUT_MSG, 452 master_node); 453 if (ret) 454 mlog(ML_ERROR, "send NEGO_TIMEOUT msg to node %d fail %d\n", 455 master_node, ret); 456 } 457 } 458 459 static int o2hb_nego_timeout_handler(struct o2net_msg *msg, u32 len, void *data, 460 void **ret_data) 461 { 462 struct o2hb_region *reg = data; 463 struct o2hb_nego_msg *nego_msg; 464 465 nego_msg = (struct o2hb_nego_msg *)msg->buf; 466 printk(KERN_NOTICE "o2hb: receive negotiate timeout message from node %d on region %s (%s).\n", 467 nego_msg->node_num, config_item_name(®->hr_item), reg->hr_dev_name); 468 if (nego_msg->node_num < O2NM_MAX_NODES) 469 set_bit(nego_msg->node_num, reg->hr_nego_node_bitmap); 470 else 471 mlog(ML_ERROR, "got nego timeout message from bad node.\n"); 472 473 return 0; 474 } 475 476 static int o2hb_nego_approve_handler(struct o2net_msg *msg, u32 len, void *data, 477 void **ret_data) 478 { 479 struct o2hb_region *reg = data; 480 481 printk(KERN_NOTICE "o2hb: negotiate timeout approved by master node on region %s (%s).\n", 482 config_item_name(®->hr_item), reg->hr_dev_name); 483 o2hb_arm_timeout(reg); 484 return 0; 485 } 486 487 static inline void o2hb_bio_wait_init(struct o2hb_bio_wait_ctxt *wc) 488 { 489 atomic_set(&wc->wc_num_reqs, 1); 490 init_completion(&wc->wc_io_complete); 491 wc->wc_error = 0; 492 } 493 494 /* Used in error paths too */ 495 static inline void o2hb_bio_wait_dec(struct o2hb_bio_wait_ctxt *wc, 496 unsigned int num) 497 { 498 /* sadly atomic_sub_and_test() isn't available on all platforms. The 499 * good news is that the fast path only completes one at a time */ 500 while(num--) { 501 if (atomic_dec_and_test(&wc->wc_num_reqs)) { 502 BUG_ON(num > 0); 503 complete(&wc->wc_io_complete); 504 } 505 } 506 } 507 508 static void o2hb_wait_on_io(struct o2hb_region *reg, 509 struct o2hb_bio_wait_ctxt *wc) 510 { 511 o2hb_bio_wait_dec(wc, 1); 512 wait_for_completion(&wc->wc_io_complete); 513 } 514 515 static void o2hb_bio_end_io(struct bio *bio) 516 { 517 struct o2hb_bio_wait_ctxt *wc = bio->bi_private; 518 519 if (bio->bi_error) { 520 mlog(ML_ERROR, "IO Error %d\n", bio->bi_error); 521 wc->wc_error = bio->bi_error; 522 } 523 524 o2hb_bio_wait_dec(wc, 1); 525 bio_put(bio); 526 } 527 528 /* Setup a Bio to cover I/O against num_slots slots starting at 529 * start_slot. */ 530 static struct bio *o2hb_setup_one_bio(struct o2hb_region *reg, 531 struct o2hb_bio_wait_ctxt *wc, 532 unsigned int *current_slot, 533 unsigned int max_slots, int op, 534 int op_flags) 535 { 536 int len, current_page; 537 unsigned int vec_len, vec_start; 538 unsigned int bits = reg->hr_block_bits; 539 unsigned int spp = reg->hr_slots_per_page; 540 unsigned int cs = *current_slot; 541 struct bio *bio; 542 struct page *page; 543 544 /* Testing has shown this allocation to take long enough under 545 * GFP_KERNEL that the local node can get fenced. It would be 546 * nicest if we could pre-allocate these bios and avoid this 547 * all together. */ 548 bio = bio_alloc(GFP_ATOMIC, 16); 549 if (!bio) { 550 mlog(ML_ERROR, "Could not alloc slots BIO!\n"); 551 bio = ERR_PTR(-ENOMEM); 552 goto bail; 553 } 554 555 /* Must put everything in 512 byte sectors for the bio... */ 556 bio->bi_iter.bi_sector = (reg->hr_start_block + cs) << (bits - 9); 557 bio->bi_bdev = reg->hr_bdev; 558 bio->bi_private = wc; 559 bio->bi_end_io = o2hb_bio_end_io; 560 bio_set_op_attrs(bio, op, op_flags); 561 562 vec_start = (cs << bits) % PAGE_SIZE; 563 while(cs < max_slots) { 564 current_page = cs / spp; 565 page = reg->hr_slot_data[current_page]; 566 567 vec_len = min(PAGE_SIZE - vec_start, 568 (max_slots-cs) * (PAGE_SIZE/spp) ); 569 570 mlog(ML_HB_BIO, "page %d, vec_len = %u, vec_start = %u\n", 571 current_page, vec_len, vec_start); 572 573 len = bio_add_page(bio, page, vec_len, vec_start); 574 if (len != vec_len) break; 575 576 cs += vec_len / (PAGE_SIZE/spp); 577 vec_start = 0; 578 } 579 580 bail: 581 *current_slot = cs; 582 return bio; 583 } 584 585 static int o2hb_read_slots(struct o2hb_region *reg, 586 unsigned int max_slots) 587 { 588 unsigned int current_slot=0; 589 int status; 590 struct o2hb_bio_wait_ctxt wc; 591 struct bio *bio; 592 593 o2hb_bio_wait_init(&wc); 594 595 while(current_slot < max_slots) { 596 bio = o2hb_setup_one_bio(reg, &wc, ¤t_slot, max_slots, 597 REQ_OP_READ, 0); 598 if (IS_ERR(bio)) { 599 status = PTR_ERR(bio); 600 mlog_errno(status); 601 goto bail_and_wait; 602 } 603 604 atomic_inc(&wc.wc_num_reqs); 605 submit_bio(bio); 606 } 607 608 status = 0; 609 610 bail_and_wait: 611 o2hb_wait_on_io(reg, &wc); 612 if (wc.wc_error && !status) 613 status = wc.wc_error; 614 615 return status; 616 } 617 618 static int o2hb_issue_node_write(struct o2hb_region *reg, 619 struct o2hb_bio_wait_ctxt *write_wc) 620 { 621 int status; 622 unsigned int slot; 623 struct bio *bio; 624 625 o2hb_bio_wait_init(write_wc); 626 627 slot = o2nm_this_node(); 628 629 bio = o2hb_setup_one_bio(reg, write_wc, &slot, slot+1, REQ_OP_WRITE, 630 WRITE_SYNC); 631 if (IS_ERR(bio)) { 632 status = PTR_ERR(bio); 633 mlog_errno(status); 634 goto bail; 635 } 636 637 atomic_inc(&write_wc->wc_num_reqs); 638 submit_bio(bio); 639 640 status = 0; 641 bail: 642 return status; 643 } 644 645 static u32 o2hb_compute_block_crc_le(struct o2hb_region *reg, 646 struct o2hb_disk_heartbeat_block *hb_block) 647 { 648 __le32 old_cksum; 649 u32 ret; 650 651 /* We want to compute the block crc with a 0 value in the 652 * hb_cksum field. Save it off here and replace after the 653 * crc. */ 654 old_cksum = hb_block->hb_cksum; 655 hb_block->hb_cksum = 0; 656 657 ret = crc32_le(0, (unsigned char *) hb_block, reg->hr_block_bytes); 658 659 hb_block->hb_cksum = old_cksum; 660 661 return ret; 662 } 663 664 static void o2hb_dump_slot(struct o2hb_disk_heartbeat_block *hb_block) 665 { 666 mlog(ML_ERROR, "Dump slot information: seq = 0x%llx, node = %u, " 667 "cksum = 0x%x, generation 0x%llx\n", 668 (long long)le64_to_cpu(hb_block->hb_seq), 669 hb_block->hb_node, le32_to_cpu(hb_block->hb_cksum), 670 (long long)le64_to_cpu(hb_block->hb_generation)); 671 } 672 673 static int o2hb_verify_crc(struct o2hb_region *reg, 674 struct o2hb_disk_heartbeat_block *hb_block) 675 { 676 u32 read, computed; 677 678 read = le32_to_cpu(hb_block->hb_cksum); 679 computed = o2hb_compute_block_crc_le(reg, hb_block); 680 681 return read == computed; 682 } 683 684 /* 685 * Compare the slot data with what we wrote in the last iteration. 686 * If the match fails, print an appropriate error message. This is to 687 * detect errors like... another node hearting on the same slot, 688 * flaky device that is losing writes, etc. 689 * Returns 1 if check succeeds, 0 otherwise. 690 */ 691 static int o2hb_check_own_slot(struct o2hb_region *reg) 692 { 693 struct o2hb_disk_slot *slot; 694 struct o2hb_disk_heartbeat_block *hb_block; 695 char *errstr; 696 697 slot = ®->hr_slots[o2nm_this_node()]; 698 /* Don't check on our 1st timestamp */ 699 if (!slot->ds_last_time) 700 return 0; 701 702 hb_block = slot->ds_raw_block; 703 if (le64_to_cpu(hb_block->hb_seq) == slot->ds_last_time && 704 le64_to_cpu(hb_block->hb_generation) == slot->ds_last_generation && 705 hb_block->hb_node == slot->ds_node_num) 706 return 1; 707 708 #define ERRSTR1 "Another node is heartbeating on device" 709 #define ERRSTR2 "Heartbeat generation mismatch on device" 710 #define ERRSTR3 "Heartbeat sequence mismatch on device" 711 712 if (hb_block->hb_node != slot->ds_node_num) 713 errstr = ERRSTR1; 714 else if (le64_to_cpu(hb_block->hb_generation) != 715 slot->ds_last_generation) 716 errstr = ERRSTR2; 717 else 718 errstr = ERRSTR3; 719 720 mlog(ML_ERROR, "%s (%s): expected(%u:0x%llx, 0x%llx), " 721 "ondisk(%u:0x%llx, 0x%llx)\n", errstr, reg->hr_dev_name, 722 slot->ds_node_num, (unsigned long long)slot->ds_last_generation, 723 (unsigned long long)slot->ds_last_time, hb_block->hb_node, 724 (unsigned long long)le64_to_cpu(hb_block->hb_generation), 725 (unsigned long long)le64_to_cpu(hb_block->hb_seq)); 726 727 return 0; 728 } 729 730 static inline void o2hb_prepare_block(struct o2hb_region *reg, 731 u64 generation) 732 { 733 int node_num; 734 u64 cputime; 735 struct o2hb_disk_slot *slot; 736 struct o2hb_disk_heartbeat_block *hb_block; 737 738 node_num = o2nm_this_node(); 739 slot = ®->hr_slots[node_num]; 740 741 hb_block = (struct o2hb_disk_heartbeat_block *)slot->ds_raw_block; 742 memset(hb_block, 0, reg->hr_block_bytes); 743 /* TODO: time stuff */ 744 cputime = CURRENT_TIME.tv_sec; 745 if (!cputime) 746 cputime = 1; 747 748 hb_block->hb_seq = cpu_to_le64(cputime); 749 hb_block->hb_node = node_num; 750 hb_block->hb_generation = cpu_to_le64(generation); 751 hb_block->hb_dead_ms = cpu_to_le32(o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS); 752 753 /* This step must always happen last! */ 754 hb_block->hb_cksum = cpu_to_le32(o2hb_compute_block_crc_le(reg, 755 hb_block)); 756 757 mlog(ML_HB_BIO, "our node generation = 0x%llx, cksum = 0x%x\n", 758 (long long)generation, 759 le32_to_cpu(hb_block->hb_cksum)); 760 } 761 762 static void o2hb_fire_callbacks(struct o2hb_callback *hbcall, 763 struct o2nm_node *node, 764 int idx) 765 { 766 struct o2hb_callback_func *f; 767 768 list_for_each_entry(f, &hbcall->list, hc_item) { 769 mlog(ML_HEARTBEAT, "calling funcs %p\n", f); 770 (f->hc_func)(node, idx, f->hc_data); 771 } 772 } 773 774 /* Will run the list in order until we process the passed event */ 775 static void o2hb_run_event_list(struct o2hb_node_event *queued_event) 776 { 777 struct o2hb_callback *hbcall; 778 struct o2hb_node_event *event; 779 780 /* Holding callback sem assures we don't alter the callback 781 * lists when doing this, and serializes ourselves with other 782 * processes wanting callbacks. */ 783 down_write(&o2hb_callback_sem); 784 785 spin_lock(&o2hb_live_lock); 786 while (!list_empty(&o2hb_node_events) 787 && !list_empty(&queued_event->hn_item)) { 788 event = list_entry(o2hb_node_events.next, 789 struct o2hb_node_event, 790 hn_item); 791 list_del_init(&event->hn_item); 792 spin_unlock(&o2hb_live_lock); 793 794 mlog(ML_HEARTBEAT, "Node %s event for %d\n", 795 event->hn_event_type == O2HB_NODE_UP_CB ? "UP" : "DOWN", 796 event->hn_node_num); 797 798 hbcall = hbcall_from_type(event->hn_event_type); 799 800 /* We should *never* have gotten on to the list with a 801 * bad type... This isn't something that we should try 802 * to recover from. */ 803 BUG_ON(IS_ERR(hbcall)); 804 805 o2hb_fire_callbacks(hbcall, event->hn_node, event->hn_node_num); 806 807 spin_lock(&o2hb_live_lock); 808 } 809 spin_unlock(&o2hb_live_lock); 810 811 up_write(&o2hb_callback_sem); 812 } 813 814 static void o2hb_queue_node_event(struct o2hb_node_event *event, 815 enum o2hb_callback_type type, 816 struct o2nm_node *node, 817 int node_num) 818 { 819 assert_spin_locked(&o2hb_live_lock); 820 821 BUG_ON((!node) && (type != O2HB_NODE_DOWN_CB)); 822 823 event->hn_event_type = type; 824 event->hn_node = node; 825 event->hn_node_num = node_num; 826 827 mlog(ML_HEARTBEAT, "Queue node %s event for node %d\n", 828 type == O2HB_NODE_UP_CB ? "UP" : "DOWN", node_num); 829 830 list_add_tail(&event->hn_item, &o2hb_node_events); 831 } 832 833 static void o2hb_shutdown_slot(struct o2hb_disk_slot *slot) 834 { 835 struct o2hb_node_event event = 836 { .hn_item = LIST_HEAD_INIT(event.hn_item), }; 837 struct o2nm_node *node; 838 int queued = 0; 839 840 node = o2nm_get_node_by_num(slot->ds_node_num); 841 if (!node) 842 return; 843 844 spin_lock(&o2hb_live_lock); 845 if (!list_empty(&slot->ds_live_item)) { 846 mlog(ML_HEARTBEAT, "Shutdown, node %d leaves region\n", 847 slot->ds_node_num); 848 849 list_del_init(&slot->ds_live_item); 850 851 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { 852 clear_bit(slot->ds_node_num, o2hb_live_node_bitmap); 853 854 o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, node, 855 slot->ds_node_num); 856 queued = 1; 857 } 858 } 859 spin_unlock(&o2hb_live_lock); 860 861 if (queued) 862 o2hb_run_event_list(&event); 863 864 o2nm_node_put(node); 865 } 866 867 static void o2hb_set_quorum_device(struct o2hb_region *reg) 868 { 869 if (!o2hb_global_heartbeat_active()) 870 return; 871 872 /* Prevent race with o2hb_heartbeat_group_drop_item() */ 873 if (kthread_should_stop()) 874 return; 875 876 /* Tag region as quorum only after thread reaches steady state */ 877 if (atomic_read(®->hr_steady_iterations) != 0) 878 return; 879 880 spin_lock(&o2hb_live_lock); 881 882 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) 883 goto unlock; 884 885 /* 886 * A region can be added to the quorum only when it sees all 887 * live nodes heartbeat on it. In other words, the region has been 888 * added to all nodes. 889 */ 890 if (memcmp(reg->hr_live_node_bitmap, o2hb_live_node_bitmap, 891 sizeof(o2hb_live_node_bitmap))) 892 goto unlock; 893 894 printk(KERN_NOTICE "o2hb: Region %s (%s) is now a quorum device\n", 895 config_item_name(®->hr_item), reg->hr_dev_name); 896 897 set_bit(reg->hr_region_num, o2hb_quorum_region_bitmap); 898 899 /* 900 * If global heartbeat active, unpin all regions if the 901 * region count > CUT_OFF 902 */ 903 if (bitmap_weight(o2hb_quorum_region_bitmap, 904 O2NM_MAX_REGIONS) > O2HB_PIN_CUT_OFF) 905 o2hb_region_unpin(NULL); 906 unlock: 907 spin_unlock(&o2hb_live_lock); 908 } 909 910 static int o2hb_check_slot(struct o2hb_region *reg, 911 struct o2hb_disk_slot *slot) 912 { 913 int changed = 0, gen_changed = 0; 914 struct o2hb_node_event event = 915 { .hn_item = LIST_HEAD_INIT(event.hn_item), }; 916 struct o2nm_node *node; 917 struct o2hb_disk_heartbeat_block *hb_block = reg->hr_tmp_block; 918 u64 cputime; 919 unsigned int dead_ms = o2hb_dead_threshold * O2HB_REGION_TIMEOUT_MS; 920 unsigned int slot_dead_ms; 921 int tmp; 922 int queued = 0; 923 924 memcpy(hb_block, slot->ds_raw_block, reg->hr_block_bytes); 925 926 /* 927 * If a node is no longer configured but is still in the livemap, we 928 * may need to clear that bit from the livemap. 929 */ 930 node = o2nm_get_node_by_num(slot->ds_node_num); 931 if (!node) { 932 spin_lock(&o2hb_live_lock); 933 tmp = test_bit(slot->ds_node_num, o2hb_live_node_bitmap); 934 spin_unlock(&o2hb_live_lock); 935 if (!tmp) 936 return 0; 937 } 938 939 if (!o2hb_verify_crc(reg, hb_block)) { 940 /* all paths from here will drop o2hb_live_lock for 941 * us. */ 942 spin_lock(&o2hb_live_lock); 943 944 /* Don't print an error on the console in this case - 945 * a freshly formatted heartbeat area will not have a 946 * crc set on it. */ 947 if (list_empty(&slot->ds_live_item)) 948 goto out; 949 950 /* The node is live but pushed out a bad crc. We 951 * consider it a transient miss but don't populate any 952 * other values as they may be junk. */ 953 mlog(ML_ERROR, "Node %d has written a bad crc to %s\n", 954 slot->ds_node_num, reg->hr_dev_name); 955 o2hb_dump_slot(hb_block); 956 957 slot->ds_equal_samples++; 958 goto fire_callbacks; 959 } 960 961 /* we don't care if these wrap.. the state transitions below 962 * clear at the right places */ 963 cputime = le64_to_cpu(hb_block->hb_seq); 964 if (slot->ds_last_time != cputime) 965 slot->ds_changed_samples++; 966 else 967 slot->ds_equal_samples++; 968 slot->ds_last_time = cputime; 969 970 /* The node changed heartbeat generations. We assume this to 971 * mean it dropped off but came back before we timed out. We 972 * want to consider it down for the time being but don't want 973 * to lose any changed_samples state we might build up to 974 * considering it live again. */ 975 if (slot->ds_last_generation != le64_to_cpu(hb_block->hb_generation)) { 976 gen_changed = 1; 977 slot->ds_equal_samples = 0; 978 mlog(ML_HEARTBEAT, "Node %d changed generation (0x%llx " 979 "to 0x%llx)\n", slot->ds_node_num, 980 (long long)slot->ds_last_generation, 981 (long long)le64_to_cpu(hb_block->hb_generation)); 982 } 983 984 slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation); 985 986 mlog(ML_HEARTBEAT, "Slot %d gen 0x%llx cksum 0x%x " 987 "seq %llu last %llu changed %u equal %u\n", 988 slot->ds_node_num, (long long)slot->ds_last_generation, 989 le32_to_cpu(hb_block->hb_cksum), 990 (unsigned long long)le64_to_cpu(hb_block->hb_seq), 991 (unsigned long long)slot->ds_last_time, slot->ds_changed_samples, 992 slot->ds_equal_samples); 993 994 spin_lock(&o2hb_live_lock); 995 996 fire_callbacks: 997 /* dead nodes only come to life after some number of 998 * changes at any time during their dead time */ 999 if (list_empty(&slot->ds_live_item) && 1000 slot->ds_changed_samples >= O2HB_LIVE_THRESHOLD) { 1001 mlog(ML_HEARTBEAT, "Node %d (id 0x%llx) joined my region\n", 1002 slot->ds_node_num, (long long)slot->ds_last_generation); 1003 1004 set_bit(slot->ds_node_num, reg->hr_live_node_bitmap); 1005 1006 /* first on the list generates a callback */ 1007 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { 1008 mlog(ML_HEARTBEAT, "o2hb: Add node %d to live nodes " 1009 "bitmap\n", slot->ds_node_num); 1010 set_bit(slot->ds_node_num, o2hb_live_node_bitmap); 1011 1012 o2hb_queue_node_event(&event, O2HB_NODE_UP_CB, node, 1013 slot->ds_node_num); 1014 1015 changed = 1; 1016 queued = 1; 1017 } 1018 1019 list_add_tail(&slot->ds_live_item, 1020 &o2hb_live_slots[slot->ds_node_num]); 1021 1022 slot->ds_equal_samples = 0; 1023 1024 /* We want to be sure that all nodes agree on the 1025 * number of milliseconds before a node will be 1026 * considered dead. The self-fencing timeout is 1027 * computed from this value, and a discrepancy might 1028 * result in heartbeat calling a node dead when it 1029 * hasn't self-fenced yet. */ 1030 slot_dead_ms = le32_to_cpu(hb_block->hb_dead_ms); 1031 if (slot_dead_ms && slot_dead_ms != dead_ms) { 1032 /* TODO: Perhaps we can fail the region here. */ 1033 mlog(ML_ERROR, "Node %d on device %s has a dead count " 1034 "of %u ms, but our count is %u ms.\n" 1035 "Please double check your configuration values " 1036 "for 'O2CB_HEARTBEAT_THRESHOLD'\n", 1037 slot->ds_node_num, reg->hr_dev_name, slot_dead_ms, 1038 dead_ms); 1039 } 1040 goto out; 1041 } 1042 1043 /* if the list is dead, we're done.. */ 1044 if (list_empty(&slot->ds_live_item)) 1045 goto out; 1046 1047 /* live nodes only go dead after enough consequtive missed 1048 * samples.. reset the missed counter whenever we see 1049 * activity */ 1050 if (slot->ds_equal_samples >= o2hb_dead_threshold || gen_changed) { 1051 mlog(ML_HEARTBEAT, "Node %d left my region\n", 1052 slot->ds_node_num); 1053 1054 clear_bit(slot->ds_node_num, reg->hr_live_node_bitmap); 1055 1056 /* last off the live_slot generates a callback */ 1057 list_del_init(&slot->ds_live_item); 1058 if (list_empty(&o2hb_live_slots[slot->ds_node_num])) { 1059 mlog(ML_HEARTBEAT, "o2hb: Remove node %d from live " 1060 "nodes bitmap\n", slot->ds_node_num); 1061 clear_bit(slot->ds_node_num, o2hb_live_node_bitmap); 1062 1063 /* node can be null */ 1064 o2hb_queue_node_event(&event, O2HB_NODE_DOWN_CB, 1065 node, slot->ds_node_num); 1066 1067 changed = 1; 1068 queued = 1; 1069 } 1070 1071 /* We don't clear this because the node is still 1072 * actually writing new blocks. */ 1073 if (!gen_changed) 1074 slot->ds_changed_samples = 0; 1075 goto out; 1076 } 1077 if (slot->ds_changed_samples) { 1078 slot->ds_changed_samples = 0; 1079 slot->ds_equal_samples = 0; 1080 } 1081 out: 1082 spin_unlock(&o2hb_live_lock); 1083 1084 if (queued) 1085 o2hb_run_event_list(&event); 1086 1087 if (node) 1088 o2nm_node_put(node); 1089 return changed; 1090 } 1091 1092 static int o2hb_highest_node(unsigned long *nodes, int numbits) 1093 { 1094 return find_last_bit(nodes, numbits); 1095 } 1096 1097 static int o2hb_do_disk_heartbeat(struct o2hb_region *reg) 1098 { 1099 int i, ret, highest_node; 1100 int membership_change = 0, own_slot_ok = 0; 1101 unsigned long configured_nodes[BITS_TO_LONGS(O2NM_MAX_NODES)]; 1102 unsigned long live_node_bitmap[BITS_TO_LONGS(O2NM_MAX_NODES)]; 1103 struct o2hb_bio_wait_ctxt write_wc; 1104 1105 ret = o2nm_configured_node_map(configured_nodes, 1106 sizeof(configured_nodes)); 1107 if (ret) { 1108 mlog_errno(ret); 1109 goto bail; 1110 } 1111 1112 /* 1113 * If a node is not configured but is in the livemap, we still need 1114 * to read the slot so as to be able to remove it from the livemap. 1115 */ 1116 o2hb_fill_node_map(live_node_bitmap, sizeof(live_node_bitmap)); 1117 i = -1; 1118 while ((i = find_next_bit(live_node_bitmap, 1119 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { 1120 set_bit(i, configured_nodes); 1121 } 1122 1123 highest_node = o2hb_highest_node(configured_nodes, O2NM_MAX_NODES); 1124 if (highest_node >= O2NM_MAX_NODES) { 1125 mlog(ML_NOTICE, "o2hb: No configured nodes found!\n"); 1126 ret = -EINVAL; 1127 goto bail; 1128 } 1129 1130 /* No sense in reading the slots of nodes that don't exist 1131 * yet. Of course, if the node definitions have holes in them 1132 * then we're reading an empty slot anyway... Consider this 1133 * best-effort. */ 1134 ret = o2hb_read_slots(reg, highest_node + 1); 1135 if (ret < 0) { 1136 mlog_errno(ret); 1137 goto bail; 1138 } 1139 1140 /* With an up to date view of the slots, we can check that no 1141 * other node has been improperly configured to heartbeat in 1142 * our slot. */ 1143 own_slot_ok = o2hb_check_own_slot(reg); 1144 1145 /* fill in the proper info for our next heartbeat */ 1146 o2hb_prepare_block(reg, reg->hr_generation); 1147 1148 ret = o2hb_issue_node_write(reg, &write_wc); 1149 if (ret < 0) { 1150 mlog_errno(ret); 1151 goto bail; 1152 } 1153 1154 i = -1; 1155 while((i = find_next_bit(configured_nodes, 1156 O2NM_MAX_NODES, i + 1)) < O2NM_MAX_NODES) { 1157 membership_change |= o2hb_check_slot(reg, ®->hr_slots[i]); 1158 } 1159 1160 /* 1161 * We have to be sure we've advertised ourselves on disk 1162 * before we can go to steady state. This ensures that 1163 * people we find in our steady state have seen us. 1164 */ 1165 o2hb_wait_on_io(reg, &write_wc); 1166 if (write_wc.wc_error) { 1167 /* Do not re-arm the write timeout on I/O error - we 1168 * can't be sure that the new block ever made it to 1169 * disk */ 1170 mlog(ML_ERROR, "Write error %d on device \"%s\"\n", 1171 write_wc.wc_error, reg->hr_dev_name); 1172 ret = write_wc.wc_error; 1173 goto bail; 1174 } 1175 1176 /* Skip disarming the timeout if own slot has stale/bad data */ 1177 if (own_slot_ok) { 1178 o2hb_set_quorum_device(reg); 1179 o2hb_arm_timeout(reg); 1180 reg->hr_last_timeout_start = jiffies; 1181 } 1182 1183 bail: 1184 /* let the person who launched us know when things are steady */ 1185 if (atomic_read(®->hr_steady_iterations) != 0) { 1186 if (!ret && own_slot_ok && !membership_change) { 1187 if (atomic_dec_and_test(®->hr_steady_iterations)) 1188 wake_up(&o2hb_steady_queue); 1189 } 1190 } 1191 1192 if (atomic_read(®->hr_steady_iterations) != 0) { 1193 if (atomic_dec_and_test(®->hr_unsteady_iterations)) { 1194 printk(KERN_NOTICE "o2hb: Unable to stabilize " 1195 "heartbeart on region %s (%s)\n", 1196 config_item_name(®->hr_item), 1197 reg->hr_dev_name); 1198 atomic_set(®->hr_steady_iterations, 0); 1199 reg->hr_aborted_start = 1; 1200 wake_up(&o2hb_steady_queue); 1201 ret = -EIO; 1202 } 1203 } 1204 1205 return ret; 1206 } 1207 1208 /* 1209 * we ride the region ref that the region dir holds. before the region 1210 * dir is removed and drops it ref it will wait to tear down this 1211 * thread. 1212 */ 1213 static int o2hb_thread(void *data) 1214 { 1215 int i, ret; 1216 struct o2hb_region *reg = data; 1217 struct o2hb_bio_wait_ctxt write_wc; 1218 ktime_t before_hb, after_hb; 1219 unsigned int elapsed_msec; 1220 1221 mlog(ML_HEARTBEAT|ML_KTHREAD, "hb thread running\n"); 1222 1223 set_user_nice(current, MIN_NICE); 1224 1225 /* Pin node */ 1226 ret = o2nm_depend_this_node(); 1227 if (ret) { 1228 mlog(ML_ERROR, "Node has been deleted, ret = %d\n", ret); 1229 reg->hr_node_deleted = 1; 1230 wake_up(&o2hb_steady_queue); 1231 return 0; 1232 } 1233 1234 while (!kthread_should_stop() && 1235 !reg->hr_unclean_stop && !reg->hr_aborted_start) { 1236 /* We track the time spent inside 1237 * o2hb_do_disk_heartbeat so that we avoid more than 1238 * hr_timeout_ms between disk writes. On busy systems 1239 * this should result in a heartbeat which is less 1240 * likely to time itself out. */ 1241 before_hb = ktime_get_real(); 1242 1243 ret = o2hb_do_disk_heartbeat(reg); 1244 reg->hr_last_hb_status = ret; 1245 1246 after_hb = ktime_get_real(); 1247 1248 elapsed_msec = (unsigned int) 1249 ktime_ms_delta(after_hb, before_hb); 1250 1251 mlog(ML_HEARTBEAT, 1252 "start = %lld, end = %lld, msec = %u, ret = %d\n", 1253 before_hb.tv64, after_hb.tv64, elapsed_msec, ret); 1254 1255 if (!kthread_should_stop() && 1256 elapsed_msec < reg->hr_timeout_ms) { 1257 /* the kthread api has blocked signals for us so no 1258 * need to record the return value. */ 1259 msleep_interruptible(reg->hr_timeout_ms - elapsed_msec); 1260 } 1261 } 1262 1263 o2hb_disarm_timeout(reg); 1264 1265 /* unclean stop is only used in very bad situation */ 1266 for(i = 0; !reg->hr_unclean_stop && i < reg->hr_blocks; i++) 1267 o2hb_shutdown_slot(®->hr_slots[i]); 1268 1269 /* Explicit down notification - avoid forcing the other nodes 1270 * to timeout on this region when we could just as easily 1271 * write a clear generation - thus indicating to them that 1272 * this node has left this region. 1273 */ 1274 if (!reg->hr_unclean_stop && !reg->hr_aborted_start) { 1275 o2hb_prepare_block(reg, 0); 1276 ret = o2hb_issue_node_write(reg, &write_wc); 1277 if (ret == 0) 1278 o2hb_wait_on_io(reg, &write_wc); 1279 else 1280 mlog_errno(ret); 1281 } 1282 1283 /* Unpin node */ 1284 o2nm_undepend_this_node(); 1285 1286 mlog(ML_HEARTBEAT|ML_KTHREAD, "o2hb thread exiting\n"); 1287 1288 return 0; 1289 } 1290 1291 #ifdef CONFIG_DEBUG_FS 1292 static int o2hb_debug_open(struct inode *inode, struct file *file) 1293 { 1294 struct o2hb_debug_buf *db = inode->i_private; 1295 struct o2hb_region *reg; 1296 unsigned long map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 1297 unsigned long lts; 1298 char *buf = NULL; 1299 int i = -1; 1300 int out = 0; 1301 1302 /* max_nodes should be the largest bitmap we pass here */ 1303 BUG_ON(sizeof(map) < db->db_size); 1304 1305 buf = kmalloc(PAGE_SIZE, GFP_KERNEL); 1306 if (!buf) 1307 goto bail; 1308 1309 switch (db->db_type) { 1310 case O2HB_DB_TYPE_LIVENODES: 1311 case O2HB_DB_TYPE_LIVEREGIONS: 1312 case O2HB_DB_TYPE_QUORUMREGIONS: 1313 case O2HB_DB_TYPE_FAILEDREGIONS: 1314 spin_lock(&o2hb_live_lock); 1315 memcpy(map, db->db_data, db->db_size); 1316 spin_unlock(&o2hb_live_lock); 1317 break; 1318 1319 case O2HB_DB_TYPE_REGION_LIVENODES: 1320 spin_lock(&o2hb_live_lock); 1321 reg = (struct o2hb_region *)db->db_data; 1322 memcpy(map, reg->hr_live_node_bitmap, db->db_size); 1323 spin_unlock(&o2hb_live_lock); 1324 break; 1325 1326 case O2HB_DB_TYPE_REGION_NUMBER: 1327 reg = (struct o2hb_region *)db->db_data; 1328 out += snprintf(buf + out, PAGE_SIZE - out, "%d\n", 1329 reg->hr_region_num); 1330 goto done; 1331 1332 case O2HB_DB_TYPE_REGION_ELAPSED_TIME: 1333 reg = (struct o2hb_region *)db->db_data; 1334 lts = reg->hr_last_timeout_start; 1335 /* If 0, it has never been set before */ 1336 if (lts) 1337 lts = jiffies_to_msecs(jiffies - lts); 1338 out += snprintf(buf + out, PAGE_SIZE - out, "%lu\n", lts); 1339 goto done; 1340 1341 case O2HB_DB_TYPE_REGION_PINNED: 1342 reg = (struct o2hb_region *)db->db_data; 1343 out += snprintf(buf + out, PAGE_SIZE - out, "%u\n", 1344 !!reg->hr_item_pinned); 1345 goto done; 1346 1347 default: 1348 goto done; 1349 } 1350 1351 while ((i = find_next_bit(map, db->db_len, i + 1)) < db->db_len) 1352 out += snprintf(buf + out, PAGE_SIZE - out, "%d ", i); 1353 out += snprintf(buf + out, PAGE_SIZE - out, "\n"); 1354 1355 done: 1356 i_size_write(inode, out); 1357 1358 file->private_data = buf; 1359 1360 return 0; 1361 bail: 1362 return -ENOMEM; 1363 } 1364 1365 static int o2hb_debug_release(struct inode *inode, struct file *file) 1366 { 1367 kfree(file->private_data); 1368 return 0; 1369 } 1370 1371 static ssize_t o2hb_debug_read(struct file *file, char __user *buf, 1372 size_t nbytes, loff_t *ppos) 1373 { 1374 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data, 1375 i_size_read(file->f_mapping->host)); 1376 } 1377 #else 1378 static int o2hb_debug_open(struct inode *inode, struct file *file) 1379 { 1380 return 0; 1381 } 1382 static int o2hb_debug_release(struct inode *inode, struct file *file) 1383 { 1384 return 0; 1385 } 1386 static ssize_t o2hb_debug_read(struct file *file, char __user *buf, 1387 size_t nbytes, loff_t *ppos) 1388 { 1389 return 0; 1390 } 1391 #endif /* CONFIG_DEBUG_FS */ 1392 1393 static const struct file_operations o2hb_debug_fops = { 1394 .open = o2hb_debug_open, 1395 .release = o2hb_debug_release, 1396 .read = o2hb_debug_read, 1397 .llseek = generic_file_llseek, 1398 }; 1399 1400 void o2hb_exit(void) 1401 { 1402 debugfs_remove(o2hb_debug_failedregions); 1403 debugfs_remove(o2hb_debug_quorumregions); 1404 debugfs_remove(o2hb_debug_liveregions); 1405 debugfs_remove(o2hb_debug_livenodes); 1406 debugfs_remove(o2hb_debug_dir); 1407 kfree(o2hb_db_livenodes); 1408 kfree(o2hb_db_liveregions); 1409 kfree(o2hb_db_quorumregions); 1410 kfree(o2hb_db_failedregions); 1411 } 1412 1413 static struct dentry *o2hb_debug_create(const char *name, struct dentry *dir, 1414 struct o2hb_debug_buf **db, int db_len, 1415 int type, int size, int len, void *data) 1416 { 1417 *db = kmalloc(db_len, GFP_KERNEL); 1418 if (!*db) 1419 return NULL; 1420 1421 (*db)->db_type = type; 1422 (*db)->db_size = size; 1423 (*db)->db_len = len; 1424 (*db)->db_data = data; 1425 1426 return debugfs_create_file(name, S_IFREG|S_IRUSR, dir, *db, 1427 &o2hb_debug_fops); 1428 } 1429 1430 static int o2hb_debug_init(void) 1431 { 1432 int ret = -ENOMEM; 1433 1434 o2hb_debug_dir = debugfs_create_dir(O2HB_DEBUG_DIR, NULL); 1435 if (!o2hb_debug_dir) { 1436 mlog_errno(ret); 1437 goto bail; 1438 } 1439 1440 o2hb_debug_livenodes = o2hb_debug_create(O2HB_DEBUG_LIVENODES, 1441 o2hb_debug_dir, 1442 &o2hb_db_livenodes, 1443 sizeof(*o2hb_db_livenodes), 1444 O2HB_DB_TYPE_LIVENODES, 1445 sizeof(o2hb_live_node_bitmap), 1446 O2NM_MAX_NODES, 1447 o2hb_live_node_bitmap); 1448 if (!o2hb_debug_livenodes) { 1449 mlog_errno(ret); 1450 goto bail; 1451 } 1452 1453 o2hb_debug_liveregions = o2hb_debug_create(O2HB_DEBUG_LIVEREGIONS, 1454 o2hb_debug_dir, 1455 &o2hb_db_liveregions, 1456 sizeof(*o2hb_db_liveregions), 1457 O2HB_DB_TYPE_LIVEREGIONS, 1458 sizeof(o2hb_live_region_bitmap), 1459 O2NM_MAX_REGIONS, 1460 o2hb_live_region_bitmap); 1461 if (!o2hb_debug_liveregions) { 1462 mlog_errno(ret); 1463 goto bail; 1464 } 1465 1466 o2hb_debug_quorumregions = 1467 o2hb_debug_create(O2HB_DEBUG_QUORUMREGIONS, 1468 o2hb_debug_dir, 1469 &o2hb_db_quorumregions, 1470 sizeof(*o2hb_db_quorumregions), 1471 O2HB_DB_TYPE_QUORUMREGIONS, 1472 sizeof(o2hb_quorum_region_bitmap), 1473 O2NM_MAX_REGIONS, 1474 o2hb_quorum_region_bitmap); 1475 if (!o2hb_debug_quorumregions) { 1476 mlog_errno(ret); 1477 goto bail; 1478 } 1479 1480 o2hb_debug_failedregions = 1481 o2hb_debug_create(O2HB_DEBUG_FAILEDREGIONS, 1482 o2hb_debug_dir, 1483 &o2hb_db_failedregions, 1484 sizeof(*o2hb_db_failedregions), 1485 O2HB_DB_TYPE_FAILEDREGIONS, 1486 sizeof(o2hb_failed_region_bitmap), 1487 O2NM_MAX_REGIONS, 1488 o2hb_failed_region_bitmap); 1489 if (!o2hb_debug_failedregions) { 1490 mlog_errno(ret); 1491 goto bail; 1492 } 1493 1494 ret = 0; 1495 bail: 1496 if (ret) 1497 o2hb_exit(); 1498 1499 return ret; 1500 } 1501 1502 int o2hb_init(void) 1503 { 1504 int i; 1505 1506 for (i = 0; i < ARRAY_SIZE(o2hb_callbacks); i++) 1507 INIT_LIST_HEAD(&o2hb_callbacks[i].list); 1508 1509 for (i = 0; i < ARRAY_SIZE(o2hb_live_slots); i++) 1510 INIT_LIST_HEAD(&o2hb_live_slots[i]); 1511 1512 INIT_LIST_HEAD(&o2hb_node_events); 1513 1514 memset(o2hb_live_node_bitmap, 0, sizeof(o2hb_live_node_bitmap)); 1515 memset(o2hb_region_bitmap, 0, sizeof(o2hb_region_bitmap)); 1516 memset(o2hb_live_region_bitmap, 0, sizeof(o2hb_live_region_bitmap)); 1517 memset(o2hb_quorum_region_bitmap, 0, sizeof(o2hb_quorum_region_bitmap)); 1518 memset(o2hb_failed_region_bitmap, 0, sizeof(o2hb_failed_region_bitmap)); 1519 1520 o2hb_dependent_users = 0; 1521 1522 return o2hb_debug_init(); 1523 } 1524 1525 /* if we're already in a callback then we're already serialized by the sem */ 1526 static void o2hb_fill_node_map_from_callback(unsigned long *map, 1527 unsigned bytes) 1528 { 1529 BUG_ON(bytes < (BITS_TO_LONGS(O2NM_MAX_NODES) * sizeof(unsigned long))); 1530 1531 memcpy(map, &o2hb_live_node_bitmap, bytes); 1532 } 1533 1534 /* 1535 * get a map of all nodes that are heartbeating in any regions 1536 */ 1537 void o2hb_fill_node_map(unsigned long *map, unsigned bytes) 1538 { 1539 /* callers want to serialize this map and callbacks so that they 1540 * can trust that they don't miss nodes coming to the party */ 1541 down_read(&o2hb_callback_sem); 1542 spin_lock(&o2hb_live_lock); 1543 o2hb_fill_node_map_from_callback(map, bytes); 1544 spin_unlock(&o2hb_live_lock); 1545 up_read(&o2hb_callback_sem); 1546 } 1547 EXPORT_SYMBOL_GPL(o2hb_fill_node_map); 1548 1549 /* 1550 * heartbeat configfs bits. The heartbeat set is a default set under 1551 * the cluster set in nodemanager.c. 1552 */ 1553 1554 static struct o2hb_region *to_o2hb_region(struct config_item *item) 1555 { 1556 return item ? container_of(item, struct o2hb_region, hr_item) : NULL; 1557 } 1558 1559 /* drop_item only drops its ref after killing the thread, nothing should 1560 * be using the region anymore. this has to clean up any state that 1561 * attributes might have built up. */ 1562 static void o2hb_region_release(struct config_item *item) 1563 { 1564 int i; 1565 struct page *page; 1566 struct o2hb_region *reg = to_o2hb_region(item); 1567 1568 mlog(ML_HEARTBEAT, "hb region release (%s)\n", reg->hr_dev_name); 1569 1570 kfree(reg->hr_tmp_block); 1571 1572 if (reg->hr_slot_data) { 1573 for (i = 0; i < reg->hr_num_pages; i++) { 1574 page = reg->hr_slot_data[i]; 1575 if (page) 1576 __free_page(page); 1577 } 1578 kfree(reg->hr_slot_data); 1579 } 1580 1581 if (reg->hr_bdev) 1582 blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE); 1583 1584 kfree(reg->hr_slots); 1585 1586 debugfs_remove(reg->hr_debug_livenodes); 1587 debugfs_remove(reg->hr_debug_regnum); 1588 debugfs_remove(reg->hr_debug_elapsed_time); 1589 debugfs_remove(reg->hr_debug_pinned); 1590 debugfs_remove(reg->hr_debug_dir); 1591 kfree(reg->hr_db_livenodes); 1592 kfree(reg->hr_db_regnum); 1593 kfree(reg->hr_db_elapsed_time); 1594 kfree(reg->hr_db_pinned); 1595 1596 spin_lock(&o2hb_live_lock); 1597 list_del(®->hr_all_item); 1598 spin_unlock(&o2hb_live_lock); 1599 1600 o2net_unregister_handler_list(®->hr_handler_list); 1601 kfree(reg); 1602 } 1603 1604 static int o2hb_read_block_input(struct o2hb_region *reg, 1605 const char *page, 1606 unsigned long *ret_bytes, 1607 unsigned int *ret_bits) 1608 { 1609 unsigned long bytes; 1610 char *p = (char *)page; 1611 1612 bytes = simple_strtoul(p, &p, 0); 1613 if (!p || (*p && (*p != '\n'))) 1614 return -EINVAL; 1615 1616 /* Heartbeat and fs min / max block sizes are the same. */ 1617 if (bytes > 4096 || bytes < 512) 1618 return -ERANGE; 1619 if (hweight16(bytes) != 1) 1620 return -EINVAL; 1621 1622 if (ret_bytes) 1623 *ret_bytes = bytes; 1624 if (ret_bits) 1625 *ret_bits = ffs(bytes) - 1; 1626 1627 return 0; 1628 } 1629 1630 static ssize_t o2hb_region_block_bytes_show(struct config_item *item, 1631 char *page) 1632 { 1633 return sprintf(page, "%u\n", to_o2hb_region(item)->hr_block_bytes); 1634 } 1635 1636 static ssize_t o2hb_region_block_bytes_store(struct config_item *item, 1637 const char *page, 1638 size_t count) 1639 { 1640 struct o2hb_region *reg = to_o2hb_region(item); 1641 int status; 1642 unsigned long block_bytes; 1643 unsigned int block_bits; 1644 1645 if (reg->hr_bdev) 1646 return -EINVAL; 1647 1648 status = o2hb_read_block_input(reg, page, &block_bytes, 1649 &block_bits); 1650 if (status) 1651 return status; 1652 1653 reg->hr_block_bytes = (unsigned int)block_bytes; 1654 reg->hr_block_bits = block_bits; 1655 1656 return count; 1657 } 1658 1659 static ssize_t o2hb_region_start_block_show(struct config_item *item, 1660 char *page) 1661 { 1662 return sprintf(page, "%llu\n", to_o2hb_region(item)->hr_start_block); 1663 } 1664 1665 static ssize_t o2hb_region_start_block_store(struct config_item *item, 1666 const char *page, 1667 size_t count) 1668 { 1669 struct o2hb_region *reg = to_o2hb_region(item); 1670 unsigned long long tmp; 1671 char *p = (char *)page; 1672 1673 if (reg->hr_bdev) 1674 return -EINVAL; 1675 1676 tmp = simple_strtoull(p, &p, 0); 1677 if (!p || (*p && (*p != '\n'))) 1678 return -EINVAL; 1679 1680 reg->hr_start_block = tmp; 1681 1682 return count; 1683 } 1684 1685 static ssize_t o2hb_region_blocks_show(struct config_item *item, char *page) 1686 { 1687 return sprintf(page, "%d\n", to_o2hb_region(item)->hr_blocks); 1688 } 1689 1690 static ssize_t o2hb_region_blocks_store(struct config_item *item, 1691 const char *page, 1692 size_t count) 1693 { 1694 struct o2hb_region *reg = to_o2hb_region(item); 1695 unsigned long tmp; 1696 char *p = (char *)page; 1697 1698 if (reg->hr_bdev) 1699 return -EINVAL; 1700 1701 tmp = simple_strtoul(p, &p, 0); 1702 if (!p || (*p && (*p != '\n'))) 1703 return -EINVAL; 1704 1705 if (tmp > O2NM_MAX_NODES || tmp == 0) 1706 return -ERANGE; 1707 1708 reg->hr_blocks = (unsigned int)tmp; 1709 1710 return count; 1711 } 1712 1713 static ssize_t o2hb_region_dev_show(struct config_item *item, char *page) 1714 { 1715 unsigned int ret = 0; 1716 1717 if (to_o2hb_region(item)->hr_bdev) 1718 ret = sprintf(page, "%s\n", to_o2hb_region(item)->hr_dev_name); 1719 1720 return ret; 1721 } 1722 1723 static void o2hb_init_region_params(struct o2hb_region *reg) 1724 { 1725 reg->hr_slots_per_page = PAGE_SIZE >> reg->hr_block_bits; 1726 reg->hr_timeout_ms = O2HB_REGION_TIMEOUT_MS; 1727 1728 mlog(ML_HEARTBEAT, "hr_start_block = %llu, hr_blocks = %u\n", 1729 reg->hr_start_block, reg->hr_blocks); 1730 mlog(ML_HEARTBEAT, "hr_block_bytes = %u, hr_block_bits = %u\n", 1731 reg->hr_block_bytes, reg->hr_block_bits); 1732 mlog(ML_HEARTBEAT, "hr_timeout_ms = %u\n", reg->hr_timeout_ms); 1733 mlog(ML_HEARTBEAT, "dead threshold = %u\n", o2hb_dead_threshold); 1734 } 1735 1736 static int o2hb_map_slot_data(struct o2hb_region *reg) 1737 { 1738 int i, j; 1739 unsigned int last_slot; 1740 unsigned int spp = reg->hr_slots_per_page; 1741 struct page *page; 1742 char *raw; 1743 struct o2hb_disk_slot *slot; 1744 1745 reg->hr_tmp_block = kmalloc(reg->hr_block_bytes, GFP_KERNEL); 1746 if (reg->hr_tmp_block == NULL) 1747 return -ENOMEM; 1748 1749 reg->hr_slots = kcalloc(reg->hr_blocks, 1750 sizeof(struct o2hb_disk_slot), GFP_KERNEL); 1751 if (reg->hr_slots == NULL) 1752 return -ENOMEM; 1753 1754 for(i = 0; i < reg->hr_blocks; i++) { 1755 slot = ®->hr_slots[i]; 1756 slot->ds_node_num = i; 1757 INIT_LIST_HEAD(&slot->ds_live_item); 1758 slot->ds_raw_block = NULL; 1759 } 1760 1761 reg->hr_num_pages = (reg->hr_blocks + spp - 1) / spp; 1762 mlog(ML_HEARTBEAT, "Going to require %u pages to cover %u blocks " 1763 "at %u blocks per page\n", 1764 reg->hr_num_pages, reg->hr_blocks, spp); 1765 1766 reg->hr_slot_data = kcalloc(reg->hr_num_pages, sizeof(struct page *), 1767 GFP_KERNEL); 1768 if (!reg->hr_slot_data) 1769 return -ENOMEM; 1770 1771 for(i = 0; i < reg->hr_num_pages; i++) { 1772 page = alloc_page(GFP_KERNEL); 1773 if (!page) 1774 return -ENOMEM; 1775 1776 reg->hr_slot_data[i] = page; 1777 1778 last_slot = i * spp; 1779 raw = page_address(page); 1780 for (j = 0; 1781 (j < spp) && ((j + last_slot) < reg->hr_blocks); 1782 j++) { 1783 BUG_ON((j + last_slot) >= reg->hr_blocks); 1784 1785 slot = ®->hr_slots[j + last_slot]; 1786 slot->ds_raw_block = 1787 (struct o2hb_disk_heartbeat_block *) raw; 1788 1789 raw += reg->hr_block_bytes; 1790 } 1791 } 1792 1793 return 0; 1794 } 1795 1796 /* Read in all the slots available and populate the tracking 1797 * structures so that we can start with a baseline idea of what's 1798 * there. */ 1799 static int o2hb_populate_slot_data(struct o2hb_region *reg) 1800 { 1801 int ret, i; 1802 struct o2hb_disk_slot *slot; 1803 struct o2hb_disk_heartbeat_block *hb_block; 1804 1805 ret = o2hb_read_slots(reg, reg->hr_blocks); 1806 if (ret) 1807 goto out; 1808 1809 /* We only want to get an idea of the values initially in each 1810 * slot, so we do no verification - o2hb_check_slot will 1811 * actually determine if each configured slot is valid and 1812 * whether any values have changed. */ 1813 for(i = 0; i < reg->hr_blocks; i++) { 1814 slot = ®->hr_slots[i]; 1815 hb_block = (struct o2hb_disk_heartbeat_block *) slot->ds_raw_block; 1816 1817 /* Only fill the values that o2hb_check_slot uses to 1818 * determine changing slots */ 1819 slot->ds_last_time = le64_to_cpu(hb_block->hb_seq); 1820 slot->ds_last_generation = le64_to_cpu(hb_block->hb_generation); 1821 } 1822 1823 out: 1824 return ret; 1825 } 1826 1827 /* this is acting as commit; we set up all of hr_bdev and hr_task or nothing */ 1828 static ssize_t o2hb_region_dev_store(struct config_item *item, 1829 const char *page, 1830 size_t count) 1831 { 1832 struct o2hb_region *reg = to_o2hb_region(item); 1833 struct task_struct *hb_task; 1834 long fd; 1835 int sectsize; 1836 char *p = (char *)page; 1837 struct fd f; 1838 struct inode *inode; 1839 ssize_t ret = -EINVAL; 1840 int live_threshold; 1841 1842 if (reg->hr_bdev) 1843 goto out; 1844 1845 /* We can't heartbeat without having had our node number 1846 * configured yet. */ 1847 if (o2nm_this_node() == O2NM_MAX_NODES) 1848 goto out; 1849 1850 fd = simple_strtol(p, &p, 0); 1851 if (!p || (*p && (*p != '\n'))) 1852 goto out; 1853 1854 if (fd < 0 || fd >= INT_MAX) 1855 goto out; 1856 1857 f = fdget(fd); 1858 if (f.file == NULL) 1859 goto out; 1860 1861 if (reg->hr_blocks == 0 || reg->hr_start_block == 0 || 1862 reg->hr_block_bytes == 0) 1863 goto out2; 1864 1865 inode = igrab(f.file->f_mapping->host); 1866 if (inode == NULL) 1867 goto out2; 1868 1869 if (!S_ISBLK(inode->i_mode)) 1870 goto out3; 1871 1872 reg->hr_bdev = I_BDEV(f.file->f_mapping->host); 1873 ret = blkdev_get(reg->hr_bdev, FMODE_WRITE | FMODE_READ, NULL); 1874 if (ret) { 1875 reg->hr_bdev = NULL; 1876 goto out3; 1877 } 1878 inode = NULL; 1879 1880 bdevname(reg->hr_bdev, reg->hr_dev_name); 1881 1882 sectsize = bdev_logical_block_size(reg->hr_bdev); 1883 if (sectsize != reg->hr_block_bytes) { 1884 mlog(ML_ERROR, 1885 "blocksize %u incorrect for device, expected %d", 1886 reg->hr_block_bytes, sectsize); 1887 ret = -EINVAL; 1888 goto out3; 1889 } 1890 1891 o2hb_init_region_params(reg); 1892 1893 /* Generation of zero is invalid */ 1894 do { 1895 get_random_bytes(®->hr_generation, 1896 sizeof(reg->hr_generation)); 1897 } while (reg->hr_generation == 0); 1898 1899 ret = o2hb_map_slot_data(reg); 1900 if (ret) { 1901 mlog_errno(ret); 1902 goto out3; 1903 } 1904 1905 ret = o2hb_populate_slot_data(reg); 1906 if (ret) { 1907 mlog_errno(ret); 1908 goto out3; 1909 } 1910 1911 INIT_DELAYED_WORK(®->hr_write_timeout_work, o2hb_write_timeout); 1912 INIT_DELAYED_WORK(®->hr_nego_timeout_work, o2hb_nego_timeout); 1913 1914 /* 1915 * A node is considered live after it has beat LIVE_THRESHOLD 1916 * times. We're not steady until we've given them a chance 1917 * _after_ our first read. 1918 * The default threshold is bare minimum so as to limit the delay 1919 * during mounts. For global heartbeat, the threshold doubled for the 1920 * first region. 1921 */ 1922 live_threshold = O2HB_LIVE_THRESHOLD; 1923 if (o2hb_global_heartbeat_active()) { 1924 spin_lock(&o2hb_live_lock); 1925 if (bitmap_weight(o2hb_region_bitmap, O2NM_MAX_REGIONS) == 1) 1926 live_threshold <<= 1; 1927 spin_unlock(&o2hb_live_lock); 1928 } 1929 ++live_threshold; 1930 atomic_set(®->hr_steady_iterations, live_threshold); 1931 /* unsteady_iterations is triple the steady_iterations */ 1932 atomic_set(®->hr_unsteady_iterations, (live_threshold * 3)); 1933 1934 hb_task = kthread_run(o2hb_thread, reg, "o2hb-%s", 1935 reg->hr_item.ci_name); 1936 if (IS_ERR(hb_task)) { 1937 ret = PTR_ERR(hb_task); 1938 mlog_errno(ret); 1939 goto out3; 1940 } 1941 1942 spin_lock(&o2hb_live_lock); 1943 reg->hr_task = hb_task; 1944 spin_unlock(&o2hb_live_lock); 1945 1946 ret = wait_event_interruptible(o2hb_steady_queue, 1947 atomic_read(®->hr_steady_iterations) == 0 || 1948 reg->hr_node_deleted); 1949 if (ret) { 1950 atomic_set(®->hr_steady_iterations, 0); 1951 reg->hr_aborted_start = 1; 1952 } 1953 1954 if (reg->hr_aborted_start) { 1955 ret = -EIO; 1956 goto out3; 1957 } 1958 1959 if (reg->hr_node_deleted) { 1960 ret = -EINVAL; 1961 goto out3; 1962 } 1963 1964 /* Ok, we were woken. Make sure it wasn't by drop_item() */ 1965 spin_lock(&o2hb_live_lock); 1966 hb_task = reg->hr_task; 1967 if (o2hb_global_heartbeat_active()) 1968 set_bit(reg->hr_region_num, o2hb_live_region_bitmap); 1969 spin_unlock(&o2hb_live_lock); 1970 1971 if (hb_task) 1972 ret = count; 1973 else 1974 ret = -EIO; 1975 1976 if (hb_task && o2hb_global_heartbeat_active()) 1977 printk(KERN_NOTICE "o2hb: Heartbeat started on region %s (%s)\n", 1978 config_item_name(®->hr_item), reg->hr_dev_name); 1979 1980 out3: 1981 iput(inode); 1982 out2: 1983 fdput(f); 1984 out: 1985 if (ret < 0) { 1986 if (reg->hr_bdev) { 1987 blkdev_put(reg->hr_bdev, FMODE_READ|FMODE_WRITE); 1988 reg->hr_bdev = NULL; 1989 } 1990 } 1991 return ret; 1992 } 1993 1994 static ssize_t o2hb_region_pid_show(struct config_item *item, char *page) 1995 { 1996 struct o2hb_region *reg = to_o2hb_region(item); 1997 pid_t pid = 0; 1998 1999 spin_lock(&o2hb_live_lock); 2000 if (reg->hr_task) 2001 pid = task_pid_nr(reg->hr_task); 2002 spin_unlock(&o2hb_live_lock); 2003 2004 if (!pid) 2005 return 0; 2006 2007 return sprintf(page, "%u\n", pid); 2008 } 2009 2010 CONFIGFS_ATTR(o2hb_region_, block_bytes); 2011 CONFIGFS_ATTR(o2hb_region_, start_block); 2012 CONFIGFS_ATTR(o2hb_region_, blocks); 2013 CONFIGFS_ATTR(o2hb_region_, dev); 2014 CONFIGFS_ATTR_RO(o2hb_region_, pid); 2015 2016 static struct configfs_attribute *o2hb_region_attrs[] = { 2017 &o2hb_region_attr_block_bytes, 2018 &o2hb_region_attr_start_block, 2019 &o2hb_region_attr_blocks, 2020 &o2hb_region_attr_dev, 2021 &o2hb_region_attr_pid, 2022 NULL, 2023 }; 2024 2025 static struct configfs_item_operations o2hb_region_item_ops = { 2026 .release = o2hb_region_release, 2027 }; 2028 2029 static struct config_item_type o2hb_region_type = { 2030 .ct_item_ops = &o2hb_region_item_ops, 2031 .ct_attrs = o2hb_region_attrs, 2032 .ct_owner = THIS_MODULE, 2033 }; 2034 2035 /* heartbeat set */ 2036 2037 struct o2hb_heartbeat_group { 2038 struct config_group hs_group; 2039 /* some stuff? */ 2040 }; 2041 2042 static struct o2hb_heartbeat_group *to_o2hb_heartbeat_group(struct config_group *group) 2043 { 2044 return group ? 2045 container_of(group, struct o2hb_heartbeat_group, hs_group) 2046 : NULL; 2047 } 2048 2049 static int o2hb_debug_region_init(struct o2hb_region *reg, struct dentry *dir) 2050 { 2051 int ret = -ENOMEM; 2052 2053 reg->hr_debug_dir = 2054 debugfs_create_dir(config_item_name(®->hr_item), dir); 2055 if (!reg->hr_debug_dir) { 2056 mlog_errno(ret); 2057 goto bail; 2058 } 2059 2060 reg->hr_debug_livenodes = 2061 o2hb_debug_create(O2HB_DEBUG_LIVENODES, 2062 reg->hr_debug_dir, 2063 &(reg->hr_db_livenodes), 2064 sizeof(*(reg->hr_db_livenodes)), 2065 O2HB_DB_TYPE_REGION_LIVENODES, 2066 sizeof(reg->hr_live_node_bitmap), 2067 O2NM_MAX_NODES, reg); 2068 if (!reg->hr_debug_livenodes) { 2069 mlog_errno(ret); 2070 goto bail; 2071 } 2072 2073 reg->hr_debug_regnum = 2074 o2hb_debug_create(O2HB_DEBUG_REGION_NUMBER, 2075 reg->hr_debug_dir, 2076 &(reg->hr_db_regnum), 2077 sizeof(*(reg->hr_db_regnum)), 2078 O2HB_DB_TYPE_REGION_NUMBER, 2079 0, O2NM_MAX_NODES, reg); 2080 if (!reg->hr_debug_regnum) { 2081 mlog_errno(ret); 2082 goto bail; 2083 } 2084 2085 reg->hr_debug_elapsed_time = 2086 o2hb_debug_create(O2HB_DEBUG_REGION_ELAPSED_TIME, 2087 reg->hr_debug_dir, 2088 &(reg->hr_db_elapsed_time), 2089 sizeof(*(reg->hr_db_elapsed_time)), 2090 O2HB_DB_TYPE_REGION_ELAPSED_TIME, 2091 0, 0, reg); 2092 if (!reg->hr_debug_elapsed_time) { 2093 mlog_errno(ret); 2094 goto bail; 2095 } 2096 2097 reg->hr_debug_pinned = 2098 o2hb_debug_create(O2HB_DEBUG_REGION_PINNED, 2099 reg->hr_debug_dir, 2100 &(reg->hr_db_pinned), 2101 sizeof(*(reg->hr_db_pinned)), 2102 O2HB_DB_TYPE_REGION_PINNED, 2103 0, 0, reg); 2104 if (!reg->hr_debug_pinned) { 2105 mlog_errno(ret); 2106 goto bail; 2107 } 2108 2109 ret = 0; 2110 bail: 2111 return ret; 2112 } 2113 2114 static struct config_item *o2hb_heartbeat_group_make_item(struct config_group *group, 2115 const char *name) 2116 { 2117 struct o2hb_region *reg = NULL; 2118 int ret; 2119 2120 reg = kzalloc(sizeof(struct o2hb_region), GFP_KERNEL); 2121 if (reg == NULL) 2122 return ERR_PTR(-ENOMEM); 2123 2124 if (strlen(name) > O2HB_MAX_REGION_NAME_LEN) { 2125 ret = -ENAMETOOLONG; 2126 goto free; 2127 } 2128 2129 spin_lock(&o2hb_live_lock); 2130 reg->hr_region_num = 0; 2131 if (o2hb_global_heartbeat_active()) { 2132 reg->hr_region_num = find_first_zero_bit(o2hb_region_bitmap, 2133 O2NM_MAX_REGIONS); 2134 if (reg->hr_region_num >= O2NM_MAX_REGIONS) { 2135 spin_unlock(&o2hb_live_lock); 2136 ret = -EFBIG; 2137 goto free; 2138 } 2139 set_bit(reg->hr_region_num, o2hb_region_bitmap); 2140 } 2141 list_add_tail(®->hr_all_item, &o2hb_all_regions); 2142 spin_unlock(&o2hb_live_lock); 2143 2144 config_item_init_type_name(®->hr_item, name, &o2hb_region_type); 2145 2146 /* this is the same way to generate msg key as dlm, for local heartbeat, 2147 * name is also the same, so make initial crc value different to avoid 2148 * message key conflict. 2149 */ 2150 reg->hr_key = crc32_le(reg->hr_region_num + O2NM_MAX_REGIONS, 2151 name, strlen(name)); 2152 INIT_LIST_HEAD(®->hr_handler_list); 2153 ret = o2net_register_handler(O2HB_NEGO_TIMEOUT_MSG, reg->hr_key, 2154 sizeof(struct o2hb_nego_msg), 2155 o2hb_nego_timeout_handler, 2156 reg, NULL, ®->hr_handler_list); 2157 if (ret) 2158 goto free; 2159 2160 ret = o2net_register_handler(O2HB_NEGO_APPROVE_MSG, reg->hr_key, 2161 sizeof(struct o2hb_nego_msg), 2162 o2hb_nego_approve_handler, 2163 reg, NULL, ®->hr_handler_list); 2164 if (ret) 2165 goto unregister_handler; 2166 2167 ret = o2hb_debug_region_init(reg, o2hb_debug_dir); 2168 if (ret) { 2169 config_item_put(®->hr_item); 2170 goto unregister_handler; 2171 } 2172 2173 return ®->hr_item; 2174 2175 unregister_handler: 2176 o2net_unregister_handler_list(®->hr_handler_list); 2177 free: 2178 kfree(reg); 2179 return ERR_PTR(ret); 2180 } 2181 2182 static void o2hb_heartbeat_group_drop_item(struct config_group *group, 2183 struct config_item *item) 2184 { 2185 struct task_struct *hb_task; 2186 struct o2hb_region *reg = to_o2hb_region(item); 2187 int quorum_region = 0; 2188 2189 /* stop the thread when the user removes the region dir */ 2190 spin_lock(&o2hb_live_lock); 2191 hb_task = reg->hr_task; 2192 reg->hr_task = NULL; 2193 reg->hr_item_dropped = 1; 2194 spin_unlock(&o2hb_live_lock); 2195 2196 if (hb_task) 2197 kthread_stop(hb_task); 2198 2199 if (o2hb_global_heartbeat_active()) { 2200 spin_lock(&o2hb_live_lock); 2201 clear_bit(reg->hr_region_num, o2hb_region_bitmap); 2202 clear_bit(reg->hr_region_num, o2hb_live_region_bitmap); 2203 if (test_bit(reg->hr_region_num, o2hb_quorum_region_bitmap)) 2204 quorum_region = 1; 2205 clear_bit(reg->hr_region_num, o2hb_quorum_region_bitmap); 2206 spin_unlock(&o2hb_live_lock); 2207 printk(KERN_NOTICE "o2hb: Heartbeat %s on region %s (%s)\n", 2208 ((atomic_read(®->hr_steady_iterations) == 0) ? 2209 "stopped" : "start aborted"), config_item_name(item), 2210 reg->hr_dev_name); 2211 } 2212 2213 /* 2214 * If we're racing a dev_write(), we need to wake them. They will 2215 * check reg->hr_task 2216 */ 2217 if (atomic_read(®->hr_steady_iterations) != 0) { 2218 reg->hr_aborted_start = 1; 2219 atomic_set(®->hr_steady_iterations, 0); 2220 wake_up(&o2hb_steady_queue); 2221 } 2222 2223 config_item_put(item); 2224 2225 if (!o2hb_global_heartbeat_active() || !quorum_region) 2226 return; 2227 2228 /* 2229 * If global heartbeat active and there are dependent users, 2230 * pin all regions if quorum region count <= CUT_OFF 2231 */ 2232 spin_lock(&o2hb_live_lock); 2233 2234 if (!o2hb_dependent_users) 2235 goto unlock; 2236 2237 if (bitmap_weight(o2hb_quorum_region_bitmap, 2238 O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF) 2239 o2hb_region_pin(NULL); 2240 2241 unlock: 2242 spin_unlock(&o2hb_live_lock); 2243 } 2244 2245 static ssize_t o2hb_heartbeat_group_threshold_show(struct config_item *item, 2246 char *page) 2247 { 2248 return sprintf(page, "%u\n", o2hb_dead_threshold); 2249 } 2250 2251 static ssize_t o2hb_heartbeat_group_threshold_store(struct config_item *item, 2252 const char *page, size_t count) 2253 { 2254 unsigned long tmp; 2255 char *p = (char *)page; 2256 2257 tmp = simple_strtoul(p, &p, 10); 2258 if (!p || (*p && (*p != '\n'))) 2259 return -EINVAL; 2260 2261 /* this will validate ranges for us. */ 2262 o2hb_dead_threshold_set((unsigned int) tmp); 2263 2264 return count; 2265 } 2266 2267 static ssize_t o2hb_heartbeat_group_mode_show(struct config_item *item, 2268 char *page) 2269 { 2270 return sprintf(page, "%s\n", 2271 o2hb_heartbeat_mode_desc[o2hb_heartbeat_mode]); 2272 } 2273 2274 static ssize_t o2hb_heartbeat_group_mode_store(struct config_item *item, 2275 const char *page, size_t count) 2276 { 2277 unsigned int i; 2278 int ret; 2279 size_t len; 2280 2281 len = (page[count - 1] == '\n') ? count - 1 : count; 2282 if (!len) 2283 return -EINVAL; 2284 2285 for (i = 0; i < O2HB_HEARTBEAT_NUM_MODES; ++i) { 2286 if (strncasecmp(page, o2hb_heartbeat_mode_desc[i], len)) 2287 continue; 2288 2289 ret = o2hb_global_heartbeat_mode_set(i); 2290 if (!ret) 2291 printk(KERN_NOTICE "o2hb: Heartbeat mode set to %s\n", 2292 o2hb_heartbeat_mode_desc[i]); 2293 return count; 2294 } 2295 2296 return -EINVAL; 2297 2298 } 2299 2300 CONFIGFS_ATTR(o2hb_heartbeat_group_, threshold); 2301 CONFIGFS_ATTR(o2hb_heartbeat_group_, mode); 2302 2303 static struct configfs_attribute *o2hb_heartbeat_group_attrs[] = { 2304 &o2hb_heartbeat_group_attr_threshold, 2305 &o2hb_heartbeat_group_attr_mode, 2306 NULL, 2307 }; 2308 2309 static struct configfs_group_operations o2hb_heartbeat_group_group_ops = { 2310 .make_item = o2hb_heartbeat_group_make_item, 2311 .drop_item = o2hb_heartbeat_group_drop_item, 2312 }; 2313 2314 static struct config_item_type o2hb_heartbeat_group_type = { 2315 .ct_group_ops = &o2hb_heartbeat_group_group_ops, 2316 .ct_attrs = o2hb_heartbeat_group_attrs, 2317 .ct_owner = THIS_MODULE, 2318 }; 2319 2320 /* this is just here to avoid touching group in heartbeat.h which the 2321 * entire damn world #includes */ 2322 struct config_group *o2hb_alloc_hb_set(void) 2323 { 2324 struct o2hb_heartbeat_group *hs = NULL; 2325 struct config_group *ret = NULL; 2326 2327 hs = kzalloc(sizeof(struct o2hb_heartbeat_group), GFP_KERNEL); 2328 if (hs == NULL) 2329 goto out; 2330 2331 config_group_init_type_name(&hs->hs_group, "heartbeat", 2332 &o2hb_heartbeat_group_type); 2333 2334 ret = &hs->hs_group; 2335 out: 2336 if (ret == NULL) 2337 kfree(hs); 2338 return ret; 2339 } 2340 2341 void o2hb_free_hb_set(struct config_group *group) 2342 { 2343 struct o2hb_heartbeat_group *hs = to_o2hb_heartbeat_group(group); 2344 kfree(hs); 2345 } 2346 2347 /* hb callback registration and issuing */ 2348 2349 static struct o2hb_callback *hbcall_from_type(enum o2hb_callback_type type) 2350 { 2351 if (type == O2HB_NUM_CB) 2352 return ERR_PTR(-EINVAL); 2353 2354 return &o2hb_callbacks[type]; 2355 } 2356 2357 void o2hb_setup_callback(struct o2hb_callback_func *hc, 2358 enum o2hb_callback_type type, 2359 o2hb_cb_func *func, 2360 void *data, 2361 int priority) 2362 { 2363 INIT_LIST_HEAD(&hc->hc_item); 2364 hc->hc_func = func; 2365 hc->hc_data = data; 2366 hc->hc_priority = priority; 2367 hc->hc_type = type; 2368 hc->hc_magic = O2HB_CB_MAGIC; 2369 } 2370 EXPORT_SYMBOL_GPL(o2hb_setup_callback); 2371 2372 /* 2373 * In local heartbeat mode, region_uuid passed matches the dlm domain name. 2374 * In global heartbeat mode, region_uuid passed is NULL. 2375 * 2376 * In local, we only pin the matching region. In global we pin all the active 2377 * regions. 2378 */ 2379 static int o2hb_region_pin(const char *region_uuid) 2380 { 2381 int ret = 0, found = 0; 2382 struct o2hb_region *reg; 2383 char *uuid; 2384 2385 assert_spin_locked(&o2hb_live_lock); 2386 2387 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { 2388 if (reg->hr_item_dropped) 2389 continue; 2390 2391 uuid = config_item_name(®->hr_item); 2392 2393 /* local heartbeat */ 2394 if (region_uuid) { 2395 if (strcmp(region_uuid, uuid)) 2396 continue; 2397 found = 1; 2398 } 2399 2400 if (reg->hr_item_pinned || reg->hr_item_dropped) 2401 goto skip_pin; 2402 2403 /* Ignore ENOENT only for local hb (userdlm domain) */ 2404 ret = o2nm_depend_item(®->hr_item); 2405 if (!ret) { 2406 mlog(ML_CLUSTER, "Pin region %s\n", uuid); 2407 reg->hr_item_pinned = 1; 2408 } else { 2409 if (ret == -ENOENT && found) 2410 ret = 0; 2411 else { 2412 mlog(ML_ERROR, "Pin region %s fails with %d\n", 2413 uuid, ret); 2414 break; 2415 } 2416 } 2417 skip_pin: 2418 if (found) 2419 break; 2420 } 2421 2422 return ret; 2423 } 2424 2425 /* 2426 * In local heartbeat mode, region_uuid passed matches the dlm domain name. 2427 * In global heartbeat mode, region_uuid passed is NULL. 2428 * 2429 * In local, we only unpin the matching region. In global we unpin all the 2430 * active regions. 2431 */ 2432 static void o2hb_region_unpin(const char *region_uuid) 2433 { 2434 struct o2hb_region *reg; 2435 char *uuid; 2436 int found = 0; 2437 2438 assert_spin_locked(&o2hb_live_lock); 2439 2440 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { 2441 if (reg->hr_item_dropped) 2442 continue; 2443 2444 uuid = config_item_name(®->hr_item); 2445 if (region_uuid) { 2446 if (strcmp(region_uuid, uuid)) 2447 continue; 2448 found = 1; 2449 } 2450 2451 if (reg->hr_item_pinned) { 2452 mlog(ML_CLUSTER, "Unpin region %s\n", uuid); 2453 o2nm_undepend_item(®->hr_item); 2454 reg->hr_item_pinned = 0; 2455 } 2456 if (found) 2457 break; 2458 } 2459 } 2460 2461 static int o2hb_region_inc_user(const char *region_uuid) 2462 { 2463 int ret = 0; 2464 2465 spin_lock(&o2hb_live_lock); 2466 2467 /* local heartbeat */ 2468 if (!o2hb_global_heartbeat_active()) { 2469 ret = o2hb_region_pin(region_uuid); 2470 goto unlock; 2471 } 2472 2473 /* 2474 * if global heartbeat active and this is the first dependent user, 2475 * pin all regions if quorum region count <= CUT_OFF 2476 */ 2477 o2hb_dependent_users++; 2478 if (o2hb_dependent_users > 1) 2479 goto unlock; 2480 2481 if (bitmap_weight(o2hb_quorum_region_bitmap, 2482 O2NM_MAX_REGIONS) <= O2HB_PIN_CUT_OFF) 2483 ret = o2hb_region_pin(NULL); 2484 2485 unlock: 2486 spin_unlock(&o2hb_live_lock); 2487 return ret; 2488 } 2489 2490 void o2hb_region_dec_user(const char *region_uuid) 2491 { 2492 spin_lock(&o2hb_live_lock); 2493 2494 /* local heartbeat */ 2495 if (!o2hb_global_heartbeat_active()) { 2496 o2hb_region_unpin(region_uuid); 2497 goto unlock; 2498 } 2499 2500 /* 2501 * if global heartbeat active and there are no dependent users, 2502 * unpin all quorum regions 2503 */ 2504 o2hb_dependent_users--; 2505 if (!o2hb_dependent_users) 2506 o2hb_region_unpin(NULL); 2507 2508 unlock: 2509 spin_unlock(&o2hb_live_lock); 2510 } 2511 2512 int o2hb_register_callback(const char *region_uuid, 2513 struct o2hb_callback_func *hc) 2514 { 2515 struct o2hb_callback_func *f; 2516 struct o2hb_callback *hbcall; 2517 int ret; 2518 2519 BUG_ON(hc->hc_magic != O2HB_CB_MAGIC); 2520 BUG_ON(!list_empty(&hc->hc_item)); 2521 2522 hbcall = hbcall_from_type(hc->hc_type); 2523 if (IS_ERR(hbcall)) { 2524 ret = PTR_ERR(hbcall); 2525 goto out; 2526 } 2527 2528 if (region_uuid) { 2529 ret = o2hb_region_inc_user(region_uuid); 2530 if (ret) { 2531 mlog_errno(ret); 2532 goto out; 2533 } 2534 } 2535 2536 down_write(&o2hb_callback_sem); 2537 2538 list_for_each_entry(f, &hbcall->list, hc_item) { 2539 if (hc->hc_priority < f->hc_priority) { 2540 list_add_tail(&hc->hc_item, &f->hc_item); 2541 break; 2542 } 2543 } 2544 if (list_empty(&hc->hc_item)) 2545 list_add_tail(&hc->hc_item, &hbcall->list); 2546 2547 up_write(&o2hb_callback_sem); 2548 ret = 0; 2549 out: 2550 mlog(ML_CLUSTER, "returning %d on behalf of %p for funcs %p\n", 2551 ret, __builtin_return_address(0), hc); 2552 return ret; 2553 } 2554 EXPORT_SYMBOL_GPL(o2hb_register_callback); 2555 2556 void o2hb_unregister_callback(const char *region_uuid, 2557 struct o2hb_callback_func *hc) 2558 { 2559 BUG_ON(hc->hc_magic != O2HB_CB_MAGIC); 2560 2561 mlog(ML_CLUSTER, "on behalf of %p for funcs %p\n", 2562 __builtin_return_address(0), hc); 2563 2564 /* XXX Can this happen _with_ a region reference? */ 2565 if (list_empty(&hc->hc_item)) 2566 return; 2567 2568 if (region_uuid) 2569 o2hb_region_dec_user(region_uuid); 2570 2571 down_write(&o2hb_callback_sem); 2572 2573 list_del_init(&hc->hc_item); 2574 2575 up_write(&o2hb_callback_sem); 2576 } 2577 EXPORT_SYMBOL_GPL(o2hb_unregister_callback); 2578 2579 int o2hb_check_node_heartbeating(u8 node_num) 2580 { 2581 unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 2582 2583 o2hb_fill_node_map(testing_map, sizeof(testing_map)); 2584 if (!test_bit(node_num, testing_map)) { 2585 mlog(ML_HEARTBEAT, 2586 "node (%u) does not have heartbeating enabled.\n", 2587 node_num); 2588 return 0; 2589 } 2590 2591 return 1; 2592 } 2593 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating); 2594 2595 int o2hb_check_node_heartbeating_no_sem(u8 node_num) 2596 { 2597 unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 2598 2599 spin_lock(&o2hb_live_lock); 2600 o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map)); 2601 spin_unlock(&o2hb_live_lock); 2602 if (!test_bit(node_num, testing_map)) { 2603 mlog(ML_HEARTBEAT, 2604 "node (%u) does not have heartbeating enabled.\n", 2605 node_num); 2606 return 0; 2607 } 2608 2609 return 1; 2610 } 2611 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_no_sem); 2612 2613 int o2hb_check_node_heartbeating_from_callback(u8 node_num) 2614 { 2615 unsigned long testing_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 2616 2617 o2hb_fill_node_map_from_callback(testing_map, sizeof(testing_map)); 2618 if (!test_bit(node_num, testing_map)) { 2619 mlog(ML_HEARTBEAT, 2620 "node (%u) does not have heartbeating enabled.\n", 2621 node_num); 2622 return 0; 2623 } 2624 2625 return 1; 2626 } 2627 EXPORT_SYMBOL_GPL(o2hb_check_node_heartbeating_from_callback); 2628 2629 /* Makes sure our local node is configured with a node number, and is 2630 * heartbeating. */ 2631 int o2hb_check_local_node_heartbeating(void) 2632 { 2633 u8 node_num; 2634 2635 /* if this node was set then we have networking */ 2636 node_num = o2nm_this_node(); 2637 if (node_num == O2NM_MAX_NODES) { 2638 mlog(ML_HEARTBEAT, "this node has not been configured.\n"); 2639 return 0; 2640 } 2641 2642 return o2hb_check_node_heartbeating(node_num); 2643 } 2644 EXPORT_SYMBOL_GPL(o2hb_check_local_node_heartbeating); 2645 2646 /* 2647 * this is just a hack until we get the plumbing which flips file systems 2648 * read only and drops the hb ref instead of killing the node dead. 2649 */ 2650 void o2hb_stop_all_regions(void) 2651 { 2652 struct o2hb_region *reg; 2653 2654 mlog(ML_ERROR, "stopping heartbeat on all active regions.\n"); 2655 2656 spin_lock(&o2hb_live_lock); 2657 2658 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) 2659 reg->hr_unclean_stop = 1; 2660 2661 spin_unlock(&o2hb_live_lock); 2662 } 2663 EXPORT_SYMBOL_GPL(o2hb_stop_all_regions); 2664 2665 int o2hb_get_all_regions(char *region_uuids, u8 max_regions) 2666 { 2667 struct o2hb_region *reg; 2668 int numregs = 0; 2669 char *p; 2670 2671 spin_lock(&o2hb_live_lock); 2672 2673 p = region_uuids; 2674 list_for_each_entry(reg, &o2hb_all_regions, hr_all_item) { 2675 if (reg->hr_item_dropped) 2676 continue; 2677 2678 mlog(0, "Region: %s\n", config_item_name(®->hr_item)); 2679 if (numregs < max_regions) { 2680 memcpy(p, config_item_name(®->hr_item), 2681 O2HB_MAX_REGION_NAME_LEN); 2682 p += O2HB_MAX_REGION_NAME_LEN; 2683 } 2684 numregs++; 2685 } 2686 2687 spin_unlock(&o2hb_live_lock); 2688 2689 return numregs; 2690 } 2691 EXPORT_SYMBOL_GPL(o2hb_get_all_regions); 2692 2693 int o2hb_global_heartbeat_active(void) 2694 { 2695 return (o2hb_heartbeat_mode == O2HB_HEARTBEAT_GLOBAL); 2696 } 2697 EXPORT_SYMBOL(o2hb_global_heartbeat_active); 2698