1 /****************************************************************************** 2 ******************************************************************************* 3 ** 4 ** Copyright (C) 2005-2011 Red Hat, Inc. All rights reserved. 5 ** 6 ** This copyrighted material is made available to anyone wishing to use, 7 ** modify, copy, or redistribute it subject to the terms and conditions 8 ** of the GNU General Public License v.2. 9 ** 10 ******************************************************************************* 11 ******************************************************************************/ 12 13 #include "dlm_internal.h" 14 #include "lockspace.h" 15 #include "member.h" 16 #include "recoverd.h" 17 #include "recover.h" 18 #include "rcom.h" 19 #include "config.h" 20 #include "lowcomms.h" 21 22 int dlm_slots_version(struct dlm_header *h) 23 { 24 if ((h->h_version & 0x0000FFFF) < DLM_HEADER_SLOTS) 25 return 0; 26 return 1; 27 } 28 29 void dlm_slot_save(struct dlm_ls *ls, struct dlm_rcom *rc, 30 struct dlm_member *memb) 31 { 32 struct rcom_config *rf = (struct rcom_config *)rc->rc_buf; 33 34 if (!dlm_slots_version(&rc->rc_header)) 35 return; 36 37 memb->slot = le16_to_cpu(rf->rf_our_slot); 38 memb->generation = le32_to_cpu(rf->rf_generation); 39 } 40 41 void dlm_slots_copy_out(struct dlm_ls *ls, struct dlm_rcom *rc) 42 { 43 struct dlm_slot *slot; 44 struct rcom_slot *ro; 45 int i; 46 47 ro = (struct rcom_slot *)(rc->rc_buf + sizeof(struct rcom_config)); 48 49 /* ls_slots array is sparse, but not rcom_slots */ 50 51 for (i = 0; i < ls->ls_slots_size; i++) { 52 slot = &ls->ls_slots[i]; 53 if (!slot->nodeid) 54 continue; 55 ro->ro_nodeid = cpu_to_le32(slot->nodeid); 56 ro->ro_slot = cpu_to_le16(slot->slot); 57 ro++; 58 } 59 } 60 61 #define SLOT_DEBUG_LINE 128 62 63 static void log_slots(struct dlm_ls *ls, uint32_t gen, int num_slots, 64 struct rcom_slot *ro0, struct dlm_slot *array, 65 int array_size) 66 { 67 char line[SLOT_DEBUG_LINE]; 68 int len = SLOT_DEBUG_LINE - 1; 69 int pos = 0; 70 int ret, i; 71 72 memset(line, 0, sizeof(line)); 73 74 if (array) { 75 for (i = 0; i < array_size; i++) { 76 if (!array[i].nodeid) 77 continue; 78 79 ret = snprintf(line + pos, len - pos, " %d:%d", 80 array[i].slot, array[i].nodeid); 81 if (ret >= len - pos) 82 break; 83 pos += ret; 84 } 85 } else if (ro0) { 86 for (i = 0; i < num_slots; i++) { 87 ret = snprintf(line + pos, len - pos, " %d:%d", 88 ro0[i].ro_slot, ro0[i].ro_nodeid); 89 if (ret >= len - pos) 90 break; 91 pos += ret; 92 } 93 } 94 95 log_rinfo(ls, "generation %u slots %d%s", gen, num_slots, line); 96 } 97 98 int dlm_slots_copy_in(struct dlm_ls *ls) 99 { 100 struct dlm_member *memb; 101 struct dlm_rcom *rc = ls->ls_recover_buf; 102 struct rcom_config *rf = (struct rcom_config *)rc->rc_buf; 103 struct rcom_slot *ro0, *ro; 104 int our_nodeid = dlm_our_nodeid(); 105 int i, num_slots; 106 uint32_t gen; 107 108 if (!dlm_slots_version(&rc->rc_header)) 109 return -1; 110 111 gen = le32_to_cpu(rf->rf_generation); 112 if (gen <= ls->ls_generation) { 113 log_error(ls, "dlm_slots_copy_in gen %u old %u", 114 gen, ls->ls_generation); 115 } 116 ls->ls_generation = gen; 117 118 num_slots = le16_to_cpu(rf->rf_num_slots); 119 if (!num_slots) 120 return -1; 121 122 ro0 = (struct rcom_slot *)(rc->rc_buf + sizeof(struct rcom_config)); 123 124 for (i = 0, ro = ro0; i < num_slots; i++, ro++) { 125 ro->ro_nodeid = le32_to_cpu(ro->ro_nodeid); 126 ro->ro_slot = le16_to_cpu(ro->ro_slot); 127 } 128 129 log_slots(ls, gen, num_slots, ro0, NULL, 0); 130 131 list_for_each_entry(memb, &ls->ls_nodes, list) { 132 for (i = 0, ro = ro0; i < num_slots; i++, ro++) { 133 if (ro->ro_nodeid != memb->nodeid) 134 continue; 135 memb->slot = ro->ro_slot; 136 memb->slot_prev = memb->slot; 137 break; 138 } 139 140 if (memb->nodeid == our_nodeid) { 141 if (ls->ls_slot && ls->ls_slot != memb->slot) { 142 log_error(ls, "dlm_slots_copy_in our slot " 143 "changed %d %d", ls->ls_slot, 144 memb->slot); 145 return -1; 146 } 147 148 if (!ls->ls_slot) 149 ls->ls_slot = memb->slot; 150 } 151 152 if (!memb->slot) { 153 log_error(ls, "dlm_slots_copy_in nodeid %d no slot", 154 memb->nodeid); 155 return -1; 156 } 157 } 158 159 return 0; 160 } 161 162 /* for any nodes that do not support slots, we will not have set memb->slot 163 in wait_status_all(), so memb->slot will remain -1, and we will not 164 assign slots or set ls_num_slots here */ 165 166 int dlm_slots_assign(struct dlm_ls *ls, int *num_slots, int *slots_size, 167 struct dlm_slot **slots_out, uint32_t *gen_out) 168 { 169 struct dlm_member *memb; 170 struct dlm_slot *array; 171 int our_nodeid = dlm_our_nodeid(); 172 int array_size, max_slots, i; 173 int need = 0; 174 int max = 0; 175 int num = 0; 176 uint32_t gen = 0; 177 178 /* our own memb struct will have slot -1 gen 0 */ 179 180 list_for_each_entry(memb, &ls->ls_nodes, list) { 181 if (memb->nodeid == our_nodeid) { 182 memb->slot = ls->ls_slot; 183 memb->generation = ls->ls_generation; 184 break; 185 } 186 } 187 188 list_for_each_entry(memb, &ls->ls_nodes, list) { 189 if (memb->generation > gen) 190 gen = memb->generation; 191 192 /* node doesn't support slots */ 193 194 if (memb->slot == -1) 195 return -1; 196 197 /* node needs a slot assigned */ 198 199 if (!memb->slot) 200 need++; 201 202 /* node has a slot assigned */ 203 204 num++; 205 206 if (!max || max < memb->slot) 207 max = memb->slot; 208 209 /* sanity check, once slot is assigned it shouldn't change */ 210 211 if (memb->slot_prev && memb->slot && memb->slot_prev != memb->slot) { 212 log_error(ls, "nodeid %d slot changed %d %d", 213 memb->nodeid, memb->slot_prev, memb->slot); 214 return -1; 215 } 216 memb->slot_prev = memb->slot; 217 } 218 219 array_size = max + need; 220 221 array = kzalloc(array_size * sizeof(struct dlm_slot), GFP_NOFS); 222 if (!array) 223 return -ENOMEM; 224 225 num = 0; 226 227 /* fill in slots (offsets) that are used */ 228 229 list_for_each_entry(memb, &ls->ls_nodes, list) { 230 if (!memb->slot) 231 continue; 232 233 if (memb->slot > array_size) { 234 log_error(ls, "invalid slot number %d", memb->slot); 235 kfree(array); 236 return -1; 237 } 238 239 array[memb->slot - 1].nodeid = memb->nodeid; 240 array[memb->slot - 1].slot = memb->slot; 241 num++; 242 } 243 244 /* assign new slots from unused offsets */ 245 246 list_for_each_entry(memb, &ls->ls_nodes, list) { 247 if (memb->slot) 248 continue; 249 250 for (i = 0; i < array_size; i++) { 251 if (array[i].nodeid) 252 continue; 253 254 memb->slot = i + 1; 255 memb->slot_prev = memb->slot; 256 array[i].nodeid = memb->nodeid; 257 array[i].slot = memb->slot; 258 num++; 259 260 if (!ls->ls_slot && memb->nodeid == our_nodeid) 261 ls->ls_slot = memb->slot; 262 break; 263 } 264 265 if (!memb->slot) { 266 log_error(ls, "no free slot found"); 267 kfree(array); 268 return -1; 269 } 270 } 271 272 gen++; 273 274 log_slots(ls, gen, num, NULL, array, array_size); 275 276 max_slots = (dlm_config.ci_buffer_size - sizeof(struct dlm_rcom) - 277 sizeof(struct rcom_config)) / sizeof(struct rcom_slot); 278 279 if (num > max_slots) { 280 log_error(ls, "num_slots %d exceeds max_slots %d", 281 num, max_slots); 282 kfree(array); 283 return -1; 284 } 285 286 *gen_out = gen; 287 *slots_out = array; 288 *slots_size = array_size; 289 *num_slots = num; 290 return 0; 291 } 292 293 static void add_ordered_member(struct dlm_ls *ls, struct dlm_member *new) 294 { 295 struct dlm_member *memb = NULL; 296 struct list_head *tmp; 297 struct list_head *newlist = &new->list; 298 struct list_head *head = &ls->ls_nodes; 299 300 list_for_each(tmp, head) { 301 memb = list_entry(tmp, struct dlm_member, list); 302 if (new->nodeid < memb->nodeid) 303 break; 304 } 305 306 if (!memb) 307 list_add_tail(newlist, head); 308 else { 309 /* FIXME: can use list macro here */ 310 newlist->prev = tmp->prev; 311 newlist->next = tmp; 312 tmp->prev->next = newlist; 313 tmp->prev = newlist; 314 } 315 } 316 317 static int dlm_add_member(struct dlm_ls *ls, struct dlm_config_node *node) 318 { 319 struct dlm_member *memb; 320 int error; 321 322 memb = kzalloc(sizeof(struct dlm_member), GFP_NOFS); 323 if (!memb) 324 return -ENOMEM; 325 326 error = dlm_lowcomms_connect_node(node->nodeid); 327 if (error < 0) { 328 kfree(memb); 329 return error; 330 } 331 332 memb->nodeid = node->nodeid; 333 memb->weight = node->weight; 334 memb->comm_seq = node->comm_seq; 335 add_ordered_member(ls, memb); 336 ls->ls_num_nodes++; 337 return 0; 338 } 339 340 static struct dlm_member *find_memb(struct list_head *head, int nodeid) 341 { 342 struct dlm_member *memb; 343 344 list_for_each_entry(memb, head, list) { 345 if (memb->nodeid == nodeid) 346 return memb; 347 } 348 return NULL; 349 } 350 351 int dlm_is_member(struct dlm_ls *ls, int nodeid) 352 { 353 if (find_memb(&ls->ls_nodes, nodeid)) 354 return 1; 355 return 0; 356 } 357 358 int dlm_is_removed(struct dlm_ls *ls, int nodeid) 359 { 360 if (find_memb(&ls->ls_nodes_gone, nodeid)) 361 return 1; 362 return 0; 363 } 364 365 static void clear_memb_list(struct list_head *head) 366 { 367 struct dlm_member *memb; 368 369 while (!list_empty(head)) { 370 memb = list_entry(head->next, struct dlm_member, list); 371 list_del(&memb->list); 372 kfree(memb); 373 } 374 } 375 376 void dlm_clear_members(struct dlm_ls *ls) 377 { 378 clear_memb_list(&ls->ls_nodes); 379 ls->ls_num_nodes = 0; 380 } 381 382 void dlm_clear_members_gone(struct dlm_ls *ls) 383 { 384 clear_memb_list(&ls->ls_nodes_gone); 385 } 386 387 static void make_member_array(struct dlm_ls *ls) 388 { 389 struct dlm_member *memb; 390 int i, w, x = 0, total = 0, all_zero = 0, *array; 391 392 kfree(ls->ls_node_array); 393 ls->ls_node_array = NULL; 394 395 list_for_each_entry(memb, &ls->ls_nodes, list) { 396 if (memb->weight) 397 total += memb->weight; 398 } 399 400 /* all nodes revert to weight of 1 if all have weight 0 */ 401 402 if (!total) { 403 total = ls->ls_num_nodes; 404 all_zero = 1; 405 } 406 407 ls->ls_total_weight = total; 408 409 array = kmalloc(sizeof(int) * total, GFP_NOFS); 410 if (!array) 411 return; 412 413 list_for_each_entry(memb, &ls->ls_nodes, list) { 414 if (!all_zero && !memb->weight) 415 continue; 416 417 if (all_zero) 418 w = 1; 419 else 420 w = memb->weight; 421 422 DLM_ASSERT(x < total, printk("total %d x %d\n", total, x);); 423 424 for (i = 0; i < w; i++) 425 array[x++] = memb->nodeid; 426 } 427 428 ls->ls_node_array = array; 429 } 430 431 /* send a status request to all members just to establish comms connections */ 432 433 static int ping_members(struct dlm_ls *ls) 434 { 435 struct dlm_member *memb; 436 int error = 0; 437 438 list_for_each_entry(memb, &ls->ls_nodes, list) { 439 error = dlm_recovery_stopped(ls); 440 if (error) 441 break; 442 error = dlm_rcom_status(ls, memb->nodeid, 0); 443 if (error) 444 break; 445 } 446 if (error) 447 log_rinfo(ls, "ping_members aborted %d last nodeid %d", 448 error, ls->ls_recover_nodeid); 449 return error; 450 } 451 452 static void dlm_lsop_recover_prep(struct dlm_ls *ls) 453 { 454 if (!ls->ls_ops || !ls->ls_ops->recover_prep) 455 return; 456 ls->ls_ops->recover_prep(ls->ls_ops_arg); 457 } 458 459 static void dlm_lsop_recover_slot(struct dlm_ls *ls, struct dlm_member *memb) 460 { 461 struct dlm_slot slot; 462 uint32_t seq; 463 int error; 464 465 if (!ls->ls_ops || !ls->ls_ops->recover_slot) 466 return; 467 468 /* if there is no comms connection with this node 469 or the present comms connection is newer 470 than the one when this member was added, then 471 we consider the node to have failed (versus 472 being removed due to dlm_release_lockspace) */ 473 474 error = dlm_comm_seq(memb->nodeid, &seq); 475 476 if (!error && seq == memb->comm_seq) 477 return; 478 479 slot.nodeid = memb->nodeid; 480 slot.slot = memb->slot; 481 482 ls->ls_ops->recover_slot(ls->ls_ops_arg, &slot); 483 } 484 485 void dlm_lsop_recover_done(struct dlm_ls *ls) 486 { 487 struct dlm_member *memb; 488 struct dlm_slot *slots; 489 int i, num; 490 491 if (!ls->ls_ops || !ls->ls_ops->recover_done) 492 return; 493 494 num = ls->ls_num_nodes; 495 496 slots = kzalloc(num * sizeof(struct dlm_slot), GFP_KERNEL); 497 if (!slots) 498 return; 499 500 i = 0; 501 list_for_each_entry(memb, &ls->ls_nodes, list) { 502 if (i == num) { 503 log_error(ls, "dlm_lsop_recover_done bad num %d", num); 504 goto out; 505 } 506 slots[i].nodeid = memb->nodeid; 507 slots[i].slot = memb->slot; 508 i++; 509 } 510 511 ls->ls_ops->recover_done(ls->ls_ops_arg, slots, num, 512 ls->ls_slot, ls->ls_generation); 513 out: 514 kfree(slots); 515 } 516 517 static struct dlm_config_node *find_config_node(struct dlm_recover *rv, 518 int nodeid) 519 { 520 int i; 521 522 for (i = 0; i < rv->nodes_count; i++) { 523 if (rv->nodes[i].nodeid == nodeid) 524 return &rv->nodes[i]; 525 } 526 return NULL; 527 } 528 529 int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out) 530 { 531 struct dlm_member *memb, *safe; 532 struct dlm_config_node *node; 533 int i, error, neg = 0, low = -1; 534 535 /* previously removed members that we've not finished removing need to 536 count as a negative change so the "neg" recovery steps will happen */ 537 538 list_for_each_entry(memb, &ls->ls_nodes_gone, list) { 539 log_rinfo(ls, "prev removed member %d", memb->nodeid); 540 neg++; 541 } 542 543 /* move departed members from ls_nodes to ls_nodes_gone */ 544 545 list_for_each_entry_safe(memb, safe, &ls->ls_nodes, list) { 546 node = find_config_node(rv, memb->nodeid); 547 if (node && !node->new) 548 continue; 549 550 if (!node) { 551 log_rinfo(ls, "remove member %d", memb->nodeid); 552 } else { 553 /* removed and re-added */ 554 log_rinfo(ls, "remove member %d comm_seq %u %u", 555 memb->nodeid, memb->comm_seq, node->comm_seq); 556 } 557 558 neg++; 559 list_move(&memb->list, &ls->ls_nodes_gone); 560 ls->ls_num_nodes--; 561 dlm_lsop_recover_slot(ls, memb); 562 } 563 564 /* add new members to ls_nodes */ 565 566 for (i = 0; i < rv->nodes_count; i++) { 567 node = &rv->nodes[i]; 568 if (dlm_is_member(ls, node->nodeid)) 569 continue; 570 dlm_add_member(ls, node); 571 log_rinfo(ls, "add member %d", node->nodeid); 572 } 573 574 list_for_each_entry(memb, &ls->ls_nodes, list) { 575 if (low == -1 || memb->nodeid < low) 576 low = memb->nodeid; 577 } 578 ls->ls_low_nodeid = low; 579 580 make_member_array(ls); 581 *neg_out = neg; 582 583 error = ping_members(ls); 584 if (!error || error == -EPROTO) { 585 /* new_lockspace() may be waiting to know if the config 586 is good or bad */ 587 ls->ls_members_result = error; 588 complete(&ls->ls_members_done); 589 } 590 591 log_rinfo(ls, "dlm_recover_members %d nodes", ls->ls_num_nodes); 592 return error; 593 } 594 595 /* Userspace guarantees that dlm_ls_stop() has completed on all nodes before 596 dlm_ls_start() is called on any of them to start the new recovery. */ 597 598 int dlm_ls_stop(struct dlm_ls *ls) 599 { 600 int new; 601 602 /* 603 * Prevent dlm_recv from being in the middle of something when we do 604 * the stop. This includes ensuring dlm_recv isn't processing a 605 * recovery message (rcom), while dlm_recoverd is aborting and 606 * resetting things from an in-progress recovery. i.e. we want 607 * dlm_recoverd to abort its recovery without worrying about dlm_recv 608 * processing an rcom at the same time. Stopping dlm_recv also makes 609 * it easy for dlm_receive_message() to check locking stopped and add a 610 * message to the requestqueue without races. 611 */ 612 613 down_write(&ls->ls_recv_active); 614 615 /* 616 * Abort any recovery that's in progress (see RECOVER_STOP, 617 * dlm_recovery_stopped()) and tell any other threads running in the 618 * dlm to quit any processing (see RUNNING, dlm_locking_stopped()). 619 */ 620 621 spin_lock(&ls->ls_recover_lock); 622 set_bit(LSFL_RECOVER_STOP, &ls->ls_flags); 623 new = test_and_clear_bit(LSFL_RUNNING, &ls->ls_flags); 624 ls->ls_recover_seq++; 625 spin_unlock(&ls->ls_recover_lock); 626 627 /* 628 * Let dlm_recv run again, now any normal messages will be saved on the 629 * requestqueue for later. 630 */ 631 632 up_write(&ls->ls_recv_active); 633 634 /* 635 * This in_recovery lock does two things: 636 * 1) Keeps this function from returning until all threads are out 637 * of locking routines and locking is truly stopped. 638 * 2) Keeps any new requests from being processed until it's unlocked 639 * when recovery is complete. 640 */ 641 642 if (new) { 643 set_bit(LSFL_RECOVER_DOWN, &ls->ls_flags); 644 wake_up_process(ls->ls_recoverd_task); 645 wait_event(ls->ls_recover_lock_wait, 646 test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags)); 647 } 648 649 /* 650 * The recoverd suspend/resume makes sure that dlm_recoverd (if 651 * running) has noticed RECOVER_STOP above and quit processing the 652 * previous recovery. 653 */ 654 655 dlm_recoverd_suspend(ls); 656 657 spin_lock(&ls->ls_recover_lock); 658 kfree(ls->ls_slots); 659 ls->ls_slots = NULL; 660 ls->ls_num_slots = 0; 661 ls->ls_slots_size = 0; 662 ls->ls_recover_status = 0; 663 spin_unlock(&ls->ls_recover_lock); 664 665 dlm_recoverd_resume(ls); 666 667 if (!ls->ls_recover_begin) 668 ls->ls_recover_begin = jiffies; 669 670 dlm_lsop_recover_prep(ls); 671 return 0; 672 } 673 674 int dlm_ls_start(struct dlm_ls *ls) 675 { 676 struct dlm_recover *rv = NULL, *rv_old; 677 struct dlm_config_node *nodes; 678 int error, count; 679 680 rv = kzalloc(sizeof(struct dlm_recover), GFP_NOFS); 681 if (!rv) 682 return -ENOMEM; 683 684 error = dlm_config_nodes(ls->ls_name, &nodes, &count); 685 if (error < 0) 686 goto fail; 687 688 spin_lock(&ls->ls_recover_lock); 689 690 /* the lockspace needs to be stopped before it can be started */ 691 692 if (!dlm_locking_stopped(ls)) { 693 spin_unlock(&ls->ls_recover_lock); 694 log_error(ls, "start ignored: lockspace running"); 695 error = -EINVAL; 696 goto fail; 697 } 698 699 rv->nodes = nodes; 700 rv->nodes_count = count; 701 rv->seq = ++ls->ls_recover_seq; 702 rv_old = ls->ls_recover_args; 703 ls->ls_recover_args = rv; 704 spin_unlock(&ls->ls_recover_lock); 705 706 if (rv_old) { 707 log_error(ls, "unused recovery %llx %d", 708 (unsigned long long)rv_old->seq, rv_old->nodes_count); 709 kfree(rv_old->nodes); 710 kfree(rv_old); 711 } 712 713 set_bit(LSFL_RECOVER_WORK, &ls->ls_flags); 714 wake_up_process(ls->ls_recoverd_task); 715 return 0; 716 717 fail: 718 kfree(rv); 719 kfree(nodes); 720 return error; 721 } 722 723