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