1 /****************************************************************************** 2 ******************************************************************************* 3 ** 4 ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. 5 ** Copyright (C) 2005-2008 Red Hat, Inc. All rights reserved. 6 ** 7 ** This copyrighted material is made available to anyone wishing to use, 8 ** modify, copy, or redistribute it subject to the terms and conditions 9 ** of the GNU General Public License v.2. 10 ** 11 ******************************************************************************* 12 ******************************************************************************/ 13 14 #include "dlm_internal.h" 15 #include "lockspace.h" 16 #include "member.h" 17 #include "lowcomms.h" 18 #include "midcomms.h" 19 #include "rcom.h" 20 #include "recover.h" 21 #include "dir.h" 22 #include "config.h" 23 #include "memory.h" 24 #include "lock.h" 25 #include "util.h" 26 27 static int rcom_response(struct dlm_ls *ls) 28 { 29 return test_bit(LSFL_RCOM_READY, &ls->ls_flags); 30 } 31 32 static int create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len, 33 struct dlm_rcom **rc_ret, struct dlm_mhandle **mh_ret) 34 { 35 struct dlm_rcom *rc; 36 struct dlm_mhandle *mh; 37 char *mb; 38 int mb_len = sizeof(struct dlm_rcom) + len; 39 40 mh = dlm_lowcomms_get_buffer(to_nodeid, mb_len, GFP_NOFS, &mb); 41 if (!mh) { 42 log_print("create_rcom to %d type %d len %d ENOBUFS", 43 to_nodeid, type, len); 44 return -ENOBUFS; 45 } 46 memset(mb, 0, mb_len); 47 48 rc = (struct dlm_rcom *) mb; 49 50 rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR); 51 rc->rc_header.h_lockspace = ls->ls_global_id; 52 rc->rc_header.h_nodeid = dlm_our_nodeid(); 53 rc->rc_header.h_length = mb_len; 54 rc->rc_header.h_cmd = DLM_RCOM; 55 56 rc->rc_type = type; 57 58 spin_lock(&ls->ls_recover_lock); 59 rc->rc_seq = ls->ls_recover_seq; 60 spin_unlock(&ls->ls_recover_lock); 61 62 *mh_ret = mh; 63 *rc_ret = rc; 64 return 0; 65 } 66 67 static void send_rcom(struct dlm_ls *ls, struct dlm_mhandle *mh, 68 struct dlm_rcom *rc) 69 { 70 dlm_rcom_out(rc); 71 dlm_lowcomms_commit_buffer(mh); 72 } 73 74 static void set_rcom_status(struct dlm_ls *ls, struct rcom_status *rs, 75 uint32_t flags) 76 { 77 rs->rs_flags = cpu_to_le32(flags); 78 } 79 80 /* When replying to a status request, a node also sends back its 81 configuration values. The requesting node then checks that the remote 82 node is configured the same way as itself. */ 83 84 static void set_rcom_config(struct dlm_ls *ls, struct rcom_config *rf, 85 uint32_t num_slots) 86 { 87 rf->rf_lvblen = cpu_to_le32(ls->ls_lvblen); 88 rf->rf_lsflags = cpu_to_le32(ls->ls_exflags); 89 90 rf->rf_our_slot = cpu_to_le16(ls->ls_slot); 91 rf->rf_num_slots = cpu_to_le16(num_slots); 92 rf->rf_generation = cpu_to_le32(ls->ls_generation); 93 } 94 95 static int check_rcom_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid) 96 { 97 struct rcom_config *rf = (struct rcom_config *) rc->rc_buf; 98 99 if ((rc->rc_header.h_version & 0xFFFF0000) != DLM_HEADER_MAJOR) { 100 log_error(ls, "version mismatch: %x nodeid %d: %x", 101 DLM_HEADER_MAJOR | DLM_HEADER_MINOR, nodeid, 102 rc->rc_header.h_version); 103 return -EPROTO; 104 } 105 106 if (le32_to_cpu(rf->rf_lvblen) != ls->ls_lvblen || 107 le32_to_cpu(rf->rf_lsflags) != ls->ls_exflags) { 108 log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x", 109 ls->ls_lvblen, ls->ls_exflags, nodeid, 110 le32_to_cpu(rf->rf_lvblen), 111 le32_to_cpu(rf->rf_lsflags)); 112 return -EPROTO; 113 } 114 return 0; 115 } 116 117 static void allow_sync_reply(struct dlm_ls *ls, uint64_t *new_seq) 118 { 119 spin_lock(&ls->ls_rcom_spin); 120 *new_seq = ++ls->ls_rcom_seq; 121 set_bit(LSFL_RCOM_WAIT, &ls->ls_flags); 122 spin_unlock(&ls->ls_rcom_spin); 123 } 124 125 static void disallow_sync_reply(struct dlm_ls *ls) 126 { 127 spin_lock(&ls->ls_rcom_spin); 128 clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags); 129 clear_bit(LSFL_RCOM_READY, &ls->ls_flags); 130 spin_unlock(&ls->ls_rcom_spin); 131 } 132 133 /* 134 * low nodeid gathers one slot value at a time from each node. 135 * it sets need_slots=0, and saves rf_our_slot returned from each 136 * rcom_config. 137 * 138 * other nodes gather all slot values at once from the low nodeid. 139 * they set need_slots=1, and ignore the rf_our_slot returned from each 140 * rcom_config. they use the rf_num_slots returned from the low 141 * node's rcom_config. 142 */ 143 144 int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags) 145 { 146 struct dlm_rcom *rc; 147 struct dlm_mhandle *mh; 148 int error = 0; 149 150 ls->ls_recover_nodeid = nodeid; 151 152 if (nodeid == dlm_our_nodeid()) { 153 rc = ls->ls_recover_buf; 154 rc->rc_result = dlm_recover_status(ls); 155 goto out; 156 } 157 158 error = create_rcom(ls, nodeid, DLM_RCOM_STATUS, 159 sizeof(struct rcom_status), &rc, &mh); 160 if (error) 161 goto out; 162 163 set_rcom_status(ls, (struct rcom_status *)rc->rc_buf, status_flags); 164 165 allow_sync_reply(ls, &rc->rc_id); 166 memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size); 167 168 send_rcom(ls, mh, rc); 169 170 error = dlm_wait_function(ls, &rcom_response); 171 disallow_sync_reply(ls); 172 if (error) 173 goto out; 174 175 rc = ls->ls_recover_buf; 176 177 if (rc->rc_result == -ESRCH) { 178 /* we pretend the remote lockspace exists with 0 status */ 179 log_debug(ls, "remote node %d not ready", nodeid); 180 rc->rc_result = 0; 181 error = 0; 182 } else { 183 error = check_rcom_config(ls, rc, nodeid); 184 } 185 186 /* the caller looks at rc_result for the remote recovery status */ 187 out: 188 return error; 189 } 190 191 static void receive_rcom_status(struct dlm_ls *ls, struct dlm_rcom *rc_in) 192 { 193 struct dlm_rcom *rc; 194 struct dlm_mhandle *mh; 195 struct rcom_status *rs; 196 uint32_t status; 197 int nodeid = rc_in->rc_header.h_nodeid; 198 int len = sizeof(struct rcom_config); 199 int num_slots = 0; 200 int error; 201 202 if (!dlm_slots_version(&rc_in->rc_header)) { 203 status = dlm_recover_status(ls); 204 goto do_create; 205 } 206 207 rs = (struct rcom_status *)rc_in->rc_buf; 208 209 if (!(rs->rs_flags & DLM_RSF_NEED_SLOTS)) { 210 status = dlm_recover_status(ls); 211 goto do_create; 212 } 213 214 spin_lock(&ls->ls_recover_lock); 215 status = ls->ls_recover_status; 216 num_slots = ls->ls_num_slots; 217 spin_unlock(&ls->ls_recover_lock); 218 len += num_slots * sizeof(struct rcom_slot); 219 220 do_create: 221 error = create_rcom(ls, nodeid, DLM_RCOM_STATUS_REPLY, 222 len, &rc, &mh); 223 if (error) 224 return; 225 226 rc->rc_id = rc_in->rc_id; 227 rc->rc_seq_reply = rc_in->rc_seq; 228 rc->rc_result = status; 229 230 set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, num_slots); 231 232 if (!num_slots) 233 goto do_send; 234 235 spin_lock(&ls->ls_recover_lock); 236 if (ls->ls_num_slots != num_slots) { 237 spin_unlock(&ls->ls_recover_lock); 238 log_debug(ls, "receive_rcom_status num_slots %d to %d", 239 num_slots, ls->ls_num_slots); 240 rc->rc_result = 0; 241 set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, 0); 242 goto do_send; 243 } 244 245 dlm_slots_copy_out(ls, rc); 246 spin_unlock(&ls->ls_recover_lock); 247 248 do_send: 249 send_rcom(ls, mh, rc); 250 } 251 252 static void receive_sync_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in) 253 { 254 spin_lock(&ls->ls_rcom_spin); 255 if (!test_bit(LSFL_RCOM_WAIT, &ls->ls_flags) || 256 rc_in->rc_id != ls->ls_rcom_seq) { 257 log_debug(ls, "reject reply %d from %d seq %llx expect %llx", 258 rc_in->rc_type, rc_in->rc_header.h_nodeid, 259 (unsigned long long)rc_in->rc_id, 260 (unsigned long long)ls->ls_rcom_seq); 261 goto out; 262 } 263 memcpy(ls->ls_recover_buf, rc_in, rc_in->rc_header.h_length); 264 set_bit(LSFL_RCOM_READY, &ls->ls_flags); 265 clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags); 266 wake_up(&ls->ls_wait_general); 267 out: 268 spin_unlock(&ls->ls_rcom_spin); 269 } 270 271 int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len) 272 { 273 struct dlm_rcom *rc; 274 struct dlm_mhandle *mh; 275 int error = 0; 276 277 ls->ls_recover_nodeid = nodeid; 278 279 error = create_rcom(ls, nodeid, DLM_RCOM_NAMES, last_len, &rc, &mh); 280 if (error) 281 goto out; 282 memcpy(rc->rc_buf, last_name, last_len); 283 284 allow_sync_reply(ls, &rc->rc_id); 285 memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size); 286 287 send_rcom(ls, mh, rc); 288 289 error = dlm_wait_function(ls, &rcom_response); 290 disallow_sync_reply(ls); 291 out: 292 return error; 293 } 294 295 static void receive_rcom_names(struct dlm_ls *ls, struct dlm_rcom *rc_in) 296 { 297 struct dlm_rcom *rc; 298 struct dlm_mhandle *mh; 299 int error, inlen, outlen, nodeid; 300 301 nodeid = rc_in->rc_header.h_nodeid; 302 inlen = rc_in->rc_header.h_length - sizeof(struct dlm_rcom); 303 outlen = dlm_config.ci_buffer_size - sizeof(struct dlm_rcom); 304 305 error = create_rcom(ls, nodeid, DLM_RCOM_NAMES_REPLY, outlen, &rc, &mh); 306 if (error) 307 return; 308 rc->rc_id = rc_in->rc_id; 309 rc->rc_seq_reply = rc_in->rc_seq; 310 311 dlm_copy_master_names(ls, rc_in->rc_buf, inlen, rc->rc_buf, outlen, 312 nodeid); 313 send_rcom(ls, mh, rc); 314 } 315 316 int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid) 317 { 318 struct dlm_rcom *rc; 319 struct dlm_mhandle *mh; 320 struct dlm_ls *ls = r->res_ls; 321 int error; 322 323 error = create_rcom(ls, dir_nodeid, DLM_RCOM_LOOKUP, r->res_length, 324 &rc, &mh); 325 if (error) 326 goto out; 327 memcpy(rc->rc_buf, r->res_name, r->res_length); 328 rc->rc_id = (unsigned long) r->res_id; 329 330 send_rcom(ls, mh, rc); 331 out: 332 return error; 333 } 334 335 int dlm_send_rcom_lookup_dump(struct dlm_rsb *r, int to_nodeid) 336 { 337 struct dlm_rcom *rc; 338 struct dlm_mhandle *mh; 339 struct dlm_ls *ls = r->res_ls; 340 int error; 341 342 error = create_rcom(ls, to_nodeid, DLM_RCOM_LOOKUP, r->res_length, 343 &rc, &mh); 344 if (error) 345 goto out; 346 memcpy(rc->rc_buf, r->res_name, r->res_length); 347 rc->rc_id = 0xFFFFFFFF; 348 349 send_rcom(ls, mh, rc); 350 out: 351 return error; 352 } 353 354 static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in) 355 { 356 struct dlm_rcom *rc; 357 struct dlm_mhandle *mh; 358 int error, ret_nodeid, nodeid = rc_in->rc_header.h_nodeid; 359 int len = rc_in->rc_header.h_length - sizeof(struct dlm_rcom); 360 361 error = create_rcom(ls, nodeid, DLM_RCOM_LOOKUP_REPLY, 0, &rc, &mh); 362 if (error) 363 return; 364 365 if (rc_in->rc_id == 0xFFFFFFFF) { 366 log_error(ls, "receive_rcom_lookup dump from %d", nodeid); 367 dlm_dump_rsb_name(ls, rc_in->rc_buf, len); 368 return; 369 } 370 371 error = dlm_master_lookup(ls, nodeid, rc_in->rc_buf, len, 372 DLM_LU_RECOVER_MASTER, &ret_nodeid, NULL); 373 if (error) 374 ret_nodeid = error; 375 rc->rc_result = ret_nodeid; 376 rc->rc_id = rc_in->rc_id; 377 rc->rc_seq_reply = rc_in->rc_seq; 378 379 send_rcom(ls, mh, rc); 380 } 381 382 static void receive_rcom_lookup_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in) 383 { 384 dlm_recover_master_reply(ls, rc_in); 385 } 386 387 static void pack_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb, 388 struct rcom_lock *rl) 389 { 390 memset(rl, 0, sizeof(*rl)); 391 392 rl->rl_ownpid = cpu_to_le32(lkb->lkb_ownpid); 393 rl->rl_lkid = cpu_to_le32(lkb->lkb_id); 394 rl->rl_exflags = cpu_to_le32(lkb->lkb_exflags); 395 rl->rl_flags = cpu_to_le32(lkb->lkb_flags); 396 rl->rl_lvbseq = cpu_to_le32(lkb->lkb_lvbseq); 397 rl->rl_rqmode = lkb->lkb_rqmode; 398 rl->rl_grmode = lkb->lkb_grmode; 399 rl->rl_status = lkb->lkb_status; 400 rl->rl_wait_type = cpu_to_le16(lkb->lkb_wait_type); 401 402 if (lkb->lkb_bastfn) 403 rl->rl_asts |= DLM_CB_BAST; 404 if (lkb->lkb_astfn) 405 rl->rl_asts |= DLM_CB_CAST; 406 407 rl->rl_namelen = cpu_to_le16(r->res_length); 408 memcpy(rl->rl_name, r->res_name, r->res_length); 409 410 /* FIXME: might we have an lvb without DLM_LKF_VALBLK set ? 411 If so, receive_rcom_lock_args() won't take this copy. */ 412 413 if (lkb->lkb_lvbptr) 414 memcpy(rl->rl_lvb, lkb->lkb_lvbptr, r->res_ls->ls_lvblen); 415 } 416 417 int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb) 418 { 419 struct dlm_ls *ls = r->res_ls; 420 struct dlm_rcom *rc; 421 struct dlm_mhandle *mh; 422 struct rcom_lock *rl; 423 int error, len = sizeof(struct rcom_lock); 424 425 if (lkb->lkb_lvbptr) 426 len += ls->ls_lvblen; 427 428 error = create_rcom(ls, r->res_nodeid, DLM_RCOM_LOCK, len, &rc, &mh); 429 if (error) 430 goto out; 431 432 rl = (struct rcom_lock *) rc->rc_buf; 433 pack_rcom_lock(r, lkb, rl); 434 rc->rc_id = (unsigned long) r; 435 436 send_rcom(ls, mh, rc); 437 out: 438 return error; 439 } 440 441 /* needs at least dlm_rcom + rcom_lock */ 442 static void receive_rcom_lock(struct dlm_ls *ls, struct dlm_rcom *rc_in) 443 { 444 struct dlm_rcom *rc; 445 struct dlm_mhandle *mh; 446 int error, nodeid = rc_in->rc_header.h_nodeid; 447 448 dlm_recover_master_copy(ls, rc_in); 449 450 error = create_rcom(ls, nodeid, DLM_RCOM_LOCK_REPLY, 451 sizeof(struct rcom_lock), &rc, &mh); 452 if (error) 453 return; 454 455 /* We send back the same rcom_lock struct we received, but 456 dlm_recover_master_copy() has filled in rl_remid and rl_result */ 457 458 memcpy(rc->rc_buf, rc_in->rc_buf, sizeof(struct rcom_lock)); 459 rc->rc_id = rc_in->rc_id; 460 rc->rc_seq_reply = rc_in->rc_seq; 461 462 send_rcom(ls, mh, rc); 463 } 464 465 /* If the lockspace doesn't exist then still send a status message 466 back; it's possible that it just doesn't have its global_id yet. */ 467 468 int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in) 469 { 470 struct dlm_rcom *rc; 471 struct rcom_config *rf; 472 struct dlm_mhandle *mh; 473 char *mb; 474 int mb_len = sizeof(struct dlm_rcom) + sizeof(struct rcom_config); 475 476 mh = dlm_lowcomms_get_buffer(nodeid, mb_len, GFP_NOFS, &mb); 477 if (!mh) 478 return -ENOBUFS; 479 memset(mb, 0, mb_len); 480 481 rc = (struct dlm_rcom *) mb; 482 483 rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR); 484 rc->rc_header.h_lockspace = rc_in->rc_header.h_lockspace; 485 rc->rc_header.h_nodeid = dlm_our_nodeid(); 486 rc->rc_header.h_length = mb_len; 487 rc->rc_header.h_cmd = DLM_RCOM; 488 489 rc->rc_type = DLM_RCOM_STATUS_REPLY; 490 rc->rc_id = rc_in->rc_id; 491 rc->rc_seq_reply = rc_in->rc_seq; 492 rc->rc_result = -ESRCH; 493 494 rf = (struct rcom_config *) rc->rc_buf; 495 rf->rf_lvblen = cpu_to_le32(~0U); 496 497 dlm_rcom_out(rc); 498 dlm_lowcomms_commit_buffer(mh); 499 500 return 0; 501 } 502 503 /* 504 * Ignore messages for stage Y before we set 505 * recover_status bit for stage X: 506 * 507 * recover_status = 0 508 * 509 * dlm_recover_members() 510 * - send nothing 511 * - recv nothing 512 * - ignore NAMES, NAMES_REPLY 513 * - ignore LOOKUP, LOOKUP_REPLY 514 * - ignore LOCK, LOCK_REPLY 515 * 516 * recover_status |= NODES 517 * 518 * dlm_recover_members_wait() 519 * 520 * dlm_recover_directory() 521 * - send NAMES 522 * - recv NAMES_REPLY 523 * - ignore LOOKUP, LOOKUP_REPLY 524 * - ignore LOCK, LOCK_REPLY 525 * 526 * recover_status |= DIR 527 * 528 * dlm_recover_directory_wait() 529 * 530 * dlm_recover_masters() 531 * - send LOOKUP 532 * - recv LOOKUP_REPLY 533 * 534 * dlm_recover_locks() 535 * - send LOCKS 536 * - recv LOCKS_REPLY 537 * 538 * recover_status |= LOCKS 539 * 540 * dlm_recover_locks_wait() 541 * 542 * recover_status |= DONE 543 */ 544 545 /* Called by dlm_recv; corresponds to dlm_receive_message() but special 546 recovery-only comms are sent through here. */ 547 548 void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid) 549 { 550 int lock_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_lock); 551 int stop, reply = 0, names = 0, lookup = 0, lock = 0; 552 uint32_t status; 553 uint64_t seq; 554 555 switch (rc->rc_type) { 556 case DLM_RCOM_STATUS_REPLY: 557 reply = 1; 558 break; 559 case DLM_RCOM_NAMES: 560 names = 1; 561 break; 562 case DLM_RCOM_NAMES_REPLY: 563 names = 1; 564 reply = 1; 565 break; 566 case DLM_RCOM_LOOKUP: 567 lookup = 1; 568 break; 569 case DLM_RCOM_LOOKUP_REPLY: 570 lookup = 1; 571 reply = 1; 572 break; 573 case DLM_RCOM_LOCK: 574 lock = 1; 575 break; 576 case DLM_RCOM_LOCK_REPLY: 577 lock = 1; 578 reply = 1; 579 break; 580 }; 581 582 spin_lock(&ls->ls_recover_lock); 583 status = ls->ls_recover_status; 584 stop = test_bit(LSFL_RECOVERY_STOP, &ls->ls_flags); 585 seq = ls->ls_recover_seq; 586 spin_unlock(&ls->ls_recover_lock); 587 588 if (stop && (rc->rc_type != DLM_RCOM_STATUS)) 589 goto ignore; 590 591 if (reply && (rc->rc_seq_reply != seq)) 592 goto ignore; 593 594 if (!(status & DLM_RS_NODES) && (names || lookup || lock)) 595 goto ignore; 596 597 if (!(status & DLM_RS_DIR) && (lookup || lock)) 598 goto ignore; 599 600 switch (rc->rc_type) { 601 case DLM_RCOM_STATUS: 602 receive_rcom_status(ls, rc); 603 break; 604 605 case DLM_RCOM_NAMES: 606 receive_rcom_names(ls, rc); 607 break; 608 609 case DLM_RCOM_LOOKUP: 610 receive_rcom_lookup(ls, rc); 611 break; 612 613 case DLM_RCOM_LOCK: 614 if (rc->rc_header.h_length < lock_size) 615 goto Eshort; 616 receive_rcom_lock(ls, rc); 617 break; 618 619 case DLM_RCOM_STATUS_REPLY: 620 receive_sync_reply(ls, rc); 621 break; 622 623 case DLM_RCOM_NAMES_REPLY: 624 receive_sync_reply(ls, rc); 625 break; 626 627 case DLM_RCOM_LOOKUP_REPLY: 628 receive_rcom_lookup_reply(ls, rc); 629 break; 630 631 case DLM_RCOM_LOCK_REPLY: 632 if (rc->rc_header.h_length < lock_size) 633 goto Eshort; 634 dlm_recover_process_copy(ls, rc); 635 break; 636 637 default: 638 log_error(ls, "receive_rcom bad type %d", rc->rc_type); 639 } 640 return; 641 642 ignore: 643 log_limit(ls, "dlm_receive_rcom ignore msg %d " 644 "from %d %llu %llu recover seq %llu sts %x gen %u", 645 rc->rc_type, 646 nodeid, 647 (unsigned long long)rc->rc_seq, 648 (unsigned long long)rc->rc_seq_reply, 649 (unsigned long long)seq, 650 status, ls->ls_generation); 651 return; 652 Eshort: 653 log_error(ls, "recovery message %d from %d is too short", 654 rc->rc_type, nodeid); 655 } 656 657