1 /* $NetBSD: dev_manager.c,v 1.1.1.3 2009/12/02 00:26:23 haad Exp $ */ 2 3 /* 4 * Copyright (C) 2002-2004 Sistina Software, Inc. All rights reserved. 5 * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved. 6 * 7 * This file is part of LVM2. 8 * 9 * This copyrighted material is made available to anyone wishing to use, 10 * modify, copy, or redistribute it subject to the terms and conditions 11 * of the GNU Lesser General Public License v.2.1. 12 * 13 * You should have received a copy of the GNU Lesser General Public License 14 * along with this program; if not, write to the Free Software Foundation, 15 * Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA 16 */ 17 18 #include "lib.h" 19 #include "str_list.h" 20 #include "dev_manager.h" 21 #include "lvm-string.h" 22 #include "fs.h" 23 #include "defaults.h" 24 #include "segtype.h" 25 #include "display.h" 26 #include "toolcontext.h" 27 #include "targets.h" 28 #include "config.h" 29 #include "filter.h" 30 #include "activate.h" 31 32 #include <limits.h> 33 #include <dirent.h> 34 35 #define MAX_TARGET_PARAMSIZE 50000 36 #define UUID_PREFIX "LVM-" 37 38 typedef enum { 39 PRELOAD, 40 ACTIVATE, 41 DEACTIVATE, 42 SUSPEND, 43 SUSPEND_WITH_LOCKFS, 44 CLEAN 45 } action_t; 46 47 struct dev_manager { 48 struct dm_pool *mem; 49 50 struct cmd_context *cmd; 51 52 void *target_state; 53 uint32_t pvmove_mirror_count; 54 int flush_required; 55 56 char *vg_name; 57 }; 58 59 struct lv_layer { 60 struct logical_volume *lv; 61 const char *old_name; 62 }; 63 64 static char *_build_dlid(struct dm_pool *mem, const char *lvid, const char *layer) 65 { 66 char *dlid; 67 size_t len; 68 69 if (!layer) 70 layer = ""; 71 72 len = sizeof(UUID_PREFIX) + sizeof(union lvid) + strlen(layer); 73 74 if (!(dlid = dm_pool_alloc(mem, len))) { 75 log_error("_build_dlid: pool allocation failed for %" PRIsize_t 76 " %s %s.", len, lvid, layer); 77 return NULL; 78 } 79 80 sprintf(dlid, UUID_PREFIX "%s%s%s", lvid, (*layer) ? "-" : "", layer); 81 82 return dlid; 83 } 84 85 char *build_dlid(struct dev_manager *dm, const char *lvid, const char *layer) 86 { 87 return _build_dlid(dm->mem, lvid, layer); 88 } 89 90 static int _read_only_lv(struct logical_volume *lv) 91 { 92 return (!(lv->vg->status & LVM_WRITE) || !(lv->status & LVM_WRITE)); 93 } 94 95 /* 96 * Low level device-layer operations. 97 */ 98 static struct dm_task *_setup_task(const char *name, const char *uuid, 99 uint32_t *event_nr, int task, 100 uint32_t major, uint32_t minor) 101 { 102 struct dm_task *dmt; 103 104 if (!(dmt = dm_task_create(task))) 105 return_NULL; 106 107 if (name) 108 dm_task_set_name(dmt, name); 109 110 if (uuid && *uuid) 111 dm_task_set_uuid(dmt, uuid); 112 113 if (event_nr) 114 dm_task_set_event_nr(dmt, *event_nr); 115 116 if (major) 117 dm_task_set_major_minor(dmt, major, minor, 1); 118 119 return dmt; 120 } 121 122 static int _info_run(const char *name, const char *dlid, struct dm_info *info, 123 uint32_t *read_ahead, int mknodes, int with_open_count, 124 int with_read_ahead, uint32_t major, uint32_t minor) 125 { 126 int r = 0; 127 struct dm_task *dmt; 128 int dmtask; 129 130 dmtask = mknodes ? DM_DEVICE_MKNODES : DM_DEVICE_INFO; 131 132 if (!(dmt = _setup_task(name, dlid, 0, dmtask, major, minor))) 133 return_0; 134 135 if (!with_open_count) 136 if (!dm_task_no_open_count(dmt)) 137 log_error("Failed to disable open_count"); 138 139 if (!dm_task_run(dmt)) 140 goto_out; 141 142 if (!dm_task_get_info(dmt, info)) 143 goto_out; 144 145 if (with_read_ahead && info->exists) { 146 if (!dm_task_get_read_ahead(dmt, read_ahead)) 147 goto_out; 148 } else if (read_ahead) 149 *read_ahead = DM_READ_AHEAD_NONE; 150 151 r = 1; 152 153 out: 154 dm_task_destroy(dmt); 155 return r; 156 } 157 158 int device_is_usable(dev_t dev) 159 { 160 struct dm_task *dmt; 161 struct dm_info info; 162 const char *name; 163 uint64_t start, length; 164 char *target_type = NULL; 165 char *params; 166 void *next = NULL; 167 int r = 0; 168 169 if (!(dmt = dm_task_create(DM_DEVICE_STATUS))) { 170 log_error("Failed to allocate dm_task struct to check dev status"); 171 return 0; 172 } 173 174 if (!dm_task_set_major_minor(dmt, MAJOR(dev), MINOR(dev), 1)) 175 goto_out; 176 177 if (!dm_task_run(dmt)) { 178 log_error("Failed to get state of mapped device"); 179 goto out; 180 } 181 182 if (!dm_task_get_info(dmt, &info)) 183 goto_out; 184 185 if (!info.exists || info.suspended) 186 goto out; 187 188 name = dm_task_get_name(dmt); 189 190 /* FIXME Also check for mirror block_on_error and mpath no paths */ 191 /* For now, we exclude all mirrors */ 192 193 do { 194 next = dm_get_next_target(dmt, next, &start, &length, 195 &target_type, ¶ms); 196 /* Skip if target type doesn't match */ 197 if (target_type && !strcmp(target_type, "mirror")) 198 goto out; 199 } while (next); 200 201 /* FIXME Also check dependencies? */ 202 203 r = 1; 204 205 out: 206 dm_task_destroy(dmt); 207 return r; 208 } 209 210 static int _info(const char *name, const char *dlid, int mknodes, 211 int with_open_count, int with_read_ahead, 212 struct dm_info *info, uint32_t *read_ahead) 213 { 214 if (!mknodes && dlid && *dlid) { 215 if (_info_run(NULL, dlid, info, read_ahead, 0, with_open_count, 216 with_read_ahead, 0, 0) && 217 info->exists) 218 return 1; 219 else if (_info_run(NULL, dlid + sizeof(UUID_PREFIX) - 1, info, 220 read_ahead, 0, with_open_count, 221 with_read_ahead, 0, 0) && 222 info->exists) 223 return 1; 224 } 225 226 if (name) 227 return _info_run(name, NULL, info, read_ahead, mknodes, 228 with_open_count, with_read_ahead, 0, 0); 229 230 return 0; 231 } 232 233 static int _info_by_dev(uint32_t major, uint32_t minor, struct dm_info *info) 234 { 235 return _info_run(NULL, NULL, info, NULL, 0, 0, 0, major, minor); 236 } 237 238 int dev_manager_info(struct dm_pool *mem, const char *name, 239 const struct logical_volume *lv, int with_mknodes, 240 int with_open_count, int with_read_ahead, 241 struct dm_info *info, uint32_t *read_ahead) 242 { 243 const char *dlid; 244 245 if (!(dlid = _build_dlid(mem, lv->lvid.s, NULL))) { 246 log_error("dlid build failed for %s", lv->name); 247 return 0; 248 } 249 250 return _info(name, dlid, with_mknodes, with_open_count, with_read_ahead, 251 info, read_ahead); 252 } 253 254 /* FIXME Interface must cope with multiple targets */ 255 static int _status_run(const char *name, const char *uuid, 256 unsigned long long *s, unsigned long long *l, 257 char **t, uint32_t t_size, char **p, uint32_t p_size) 258 { 259 int r = 0; 260 struct dm_task *dmt; 261 struct dm_info info; 262 void *next = NULL; 263 uint64_t start, length; 264 char *type = NULL; 265 char *params = NULL; 266 267 if (!(dmt = _setup_task(name, uuid, 0, DM_DEVICE_STATUS, 0, 0))) 268 return_0; 269 270 if (!dm_task_no_open_count(dmt)) 271 log_error("Failed to disable open_count"); 272 273 if (!dm_task_run(dmt)) 274 goto_out; 275 276 if (!dm_task_get_info(dmt, &info) || !info.exists) 277 goto_out; 278 279 do { 280 next = dm_get_next_target(dmt, next, &start, &length, 281 &type, ¶ms); 282 if (type) { 283 *s = start; 284 *l = length; 285 /* Make sure things are null terminated */ 286 strncpy(*t, type, t_size); 287 (*t)[t_size - 1] = '\0'; 288 strncpy(*p, params, p_size); 289 (*p)[p_size - 1] = '\0'; 290 291 r = 1; 292 /* FIXME Cope with multiple targets! */ 293 break; 294 } 295 296 } while (next); 297 298 out: 299 dm_task_destroy(dmt); 300 return r; 301 } 302 303 static int _status(const char *name, const char *uuid, 304 unsigned long long *start, unsigned long long *length, 305 char **type, uint32_t type_size, char **params, 306 uint32_t param_size) __attribute__ ((unused)); 307 308 static int _status(const char *name, const char *uuid, 309 unsigned long long *start, unsigned long long *length, 310 char **type, uint32_t type_size, char **params, 311 uint32_t param_size) 312 { 313 if (uuid && *uuid) { 314 if (_status_run(NULL, uuid, start, length, type, 315 type_size, params, param_size) && 316 *params) 317 return 1; 318 else if (_status_run(NULL, uuid + sizeof(UUID_PREFIX) - 1, start, 319 length, type, type_size, params, 320 param_size) && 321 *params) 322 return 1; 323 } 324 325 if (name && _status_run(name, NULL, start, length, type, type_size, 326 params, param_size)) 327 return 1; 328 329 return 0; 330 } 331 332 static percent_range_t _combine_percent_ranges(percent_range_t a, 333 percent_range_t b) 334 { 335 if (a == PERCENT_INVALID || b == PERCENT_INVALID) 336 return PERCENT_INVALID; 337 338 if (a == PERCENT_100 && b == PERCENT_100) 339 return PERCENT_100; 340 341 if (a == PERCENT_0 && b == PERCENT_0) 342 return PERCENT_0; 343 344 return PERCENT_0_TO_100; 345 } 346 347 static int _percent_run(struct dev_manager *dm, const char *name, 348 const char *dlid, 349 const char *target_type, int wait, 350 struct logical_volume *lv, float *percent, 351 percent_range_t *overall_percent_range, 352 uint32_t *event_nr) 353 { 354 int r = 0; 355 struct dm_task *dmt; 356 struct dm_info info; 357 void *next = NULL; 358 uint64_t start, length; 359 char *type = NULL; 360 char *params = NULL; 361 struct dm_list *segh = &lv->segments; 362 struct lv_segment *seg = NULL; 363 struct segment_type *segtype; 364 percent_range_t percent_range = 0, combined_percent_range = 0; 365 int first_time = 1; 366 367 uint64_t total_numerator = 0, total_denominator = 0; 368 369 *percent = -1; 370 *overall_percent_range = PERCENT_INVALID; 371 372 if (!(dmt = _setup_task(name, dlid, event_nr, 373 wait ? DM_DEVICE_WAITEVENT : DM_DEVICE_STATUS, 0, 0))) 374 return_0; 375 376 if (!dm_task_no_open_count(dmt)) 377 log_error("Failed to disable open_count"); 378 379 if (!dm_task_run(dmt)) 380 goto_out; 381 382 if (!dm_task_get_info(dmt, &info) || !info.exists) 383 goto_out; 384 385 if (event_nr) 386 *event_nr = info.event_nr; 387 388 do { 389 next = dm_get_next_target(dmt, next, &start, &length, &type, 390 ¶ms); 391 if (lv) { 392 if (!(segh = dm_list_next(&lv->segments, segh))) { 393 log_error("Number of segments in active LV %s " 394 "does not match metadata", lv->name); 395 goto out; 396 } 397 seg = dm_list_item(segh, struct lv_segment); 398 } 399 400 if (!type || !params || strcmp(type, target_type)) 401 continue; 402 403 if (!(segtype = get_segtype_from_string(dm->cmd, type))) 404 continue; 405 406 if (segtype->ops->target_percent && 407 !segtype->ops->target_percent(&dm->target_state, 408 &percent_range, dm->mem, 409 dm->cmd, seg, params, 410 &total_numerator, 411 &total_denominator)) 412 goto_out; 413 414 if (first_time) { 415 combined_percent_range = percent_range; 416 first_time = 0; 417 } else 418 combined_percent_range = 419 _combine_percent_ranges(combined_percent_range, 420 percent_range); 421 } while (next); 422 423 if (lv && (segh = dm_list_next(&lv->segments, segh))) { 424 log_error("Number of segments in active LV %s does not " 425 "match metadata", lv->name); 426 goto out; 427 } 428 429 if (total_denominator) { 430 *percent = (float) total_numerator *100 / total_denominator; 431 *overall_percent_range = combined_percent_range; 432 } else { 433 *percent = 100; 434 if (first_time) 435 *overall_percent_range = PERCENT_100; 436 else 437 *overall_percent_range = combined_percent_range; 438 } 439 440 log_debug("LV percent: %f", *percent); 441 r = 1; 442 443 out: 444 dm_task_destroy(dmt); 445 return r; 446 } 447 448 static int _percent(struct dev_manager *dm, const char *name, const char *dlid, 449 const char *target_type, int wait, 450 struct logical_volume *lv, float *percent, 451 percent_range_t *overall_percent_range, uint32_t *event_nr) 452 { 453 if (dlid && *dlid) { 454 if (_percent_run(dm, NULL, dlid, target_type, wait, lv, percent, 455 overall_percent_range, event_nr)) 456 return 1; 457 else if (_percent_run(dm, NULL, dlid + sizeof(UUID_PREFIX) - 1, 458 target_type, wait, lv, percent, 459 overall_percent_range, event_nr)) 460 return 1; 461 } 462 463 if (name && _percent_run(dm, name, NULL, target_type, wait, lv, percent, 464 overall_percent_range, event_nr)) 465 return 1; 466 467 return 0; 468 } 469 470 /* 471 * dev_manager implementation. 472 */ 473 struct dev_manager *dev_manager_create(struct cmd_context *cmd, 474 const char *vg_name) 475 { 476 struct dm_pool *mem; 477 struct dev_manager *dm; 478 479 if (!(mem = dm_pool_create("dev_manager", 16 * 1024))) 480 return_NULL; 481 482 if (!(dm = dm_pool_zalloc(mem, sizeof(*dm)))) 483 goto_bad; 484 485 dm->cmd = cmd; 486 dm->mem = mem; 487 488 if (!(dm->vg_name = dm_pool_strdup(dm->mem, vg_name))) 489 goto_bad; 490 491 dm->target_state = NULL; 492 493 dm_udev_set_sync_support(cmd->current_settings.udev_sync); 494 495 return dm; 496 497 bad: 498 dm_pool_destroy(mem); 499 return NULL; 500 } 501 502 void dev_manager_destroy(struct dev_manager *dm) 503 { 504 dm_pool_destroy(dm->mem); 505 } 506 507 void dev_manager_release(void) 508 { 509 dm_lib_release(); 510 } 511 512 void dev_manager_exit(void) 513 { 514 dm_lib_exit(); 515 } 516 517 int dev_manager_snapshot_percent(struct dev_manager *dm, 518 const struct logical_volume *lv, 519 float *percent, percent_range_t *percent_range) 520 { 521 char *name; 522 const char *dlid; 523 524 /* 525 * Build a name for the top layer. 526 */ 527 if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, NULL))) 528 return_0; 529 530 if (!(dlid = build_dlid(dm, lv->lvid.s, NULL))) 531 return_0; 532 533 /* 534 * Try and get some info on this device. 535 */ 536 log_debug("Getting device status percentage for %s", name); 537 if (!(_percent(dm, name, dlid, "snapshot", 0, NULL, percent, 538 percent_range, NULL))) 539 return_0; 540 541 /* FIXME dm_pool_free ? */ 542 543 /* If the snapshot isn't available, percent will be -1 */ 544 return 1; 545 } 546 547 /* FIXME Merge with snapshot_percent, auto-detecting target type */ 548 /* FIXME Cope with more than one target */ 549 int dev_manager_mirror_percent(struct dev_manager *dm, 550 struct logical_volume *lv, int wait, 551 float *percent, percent_range_t *percent_range, 552 uint32_t *event_nr) 553 { 554 char *name; 555 const char *dlid; 556 const char *suffix = (lv_is_origin(lv)) ? "real" : NULL; 557 558 /* 559 * Build a name for the top layer. 560 */ 561 if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, suffix))) 562 return_0; 563 564 /* FIXME dm_pool_free ? */ 565 566 if (!(dlid = build_dlid(dm, lv->lvid.s, suffix))) { 567 log_error("dlid build failed for %s", lv->name); 568 return 0; 569 } 570 571 log_debug("Getting device mirror status percentage for %s", name); 572 if (!(_percent(dm, name, dlid, "mirror", wait, lv, percent, 573 percent_range, event_nr))) 574 return_0; 575 576 return 1; 577 } 578 579 #if 0 580 log_very_verbose("%s %s", sus ? "Suspending" : "Resuming", name); 581 582 log_verbose("Loading %s", dl->name); 583 log_very_verbose("Activating %s read-only", dl->name); 584 log_very_verbose("Activated %s %s %03u:%03u", dl->name, 585 dl->dlid, dl->info.major, dl->info.minor); 586 587 if (_get_flag(dl, VISIBLE)) 588 log_verbose("Removing %s", dl->name); 589 else 590 log_very_verbose("Removing %s", dl->name); 591 592 log_debug("Adding target: %" PRIu64 " %" PRIu64 " %s %s", 593 extent_size * seg->le, extent_size * seg->len, target, params); 594 595 log_debug("Adding target: 0 %" PRIu64 " snapshot-origin %s", 596 dl->lv->size, params); 597 log_debug("Adding target: 0 %" PRIu64 " snapshot %s", size, params); 598 log_debug("Getting device info for %s", dl->name); 599 600 /* Rename? */ 601 if ((suffix = strrchr(dl->dlid + sizeof(UUID_PREFIX) - 1, '-'))) 602 suffix++; 603 new_name = build_dm_name(dm->mem, dm->vg_name, dl->lv->name, 604 suffix); 605 606 static int _belong_to_vg(const char *vgname, const char *name) 607 { 608 const char *v = vgname, *n = name; 609 610 while (*v) { 611 if ((*v != *n) || (*v == '-' && *(++n) != '-')) 612 return 0; 613 v++, n++; 614 } 615 616 if (*n == '-' && *(n + 1) != '-') 617 return 1; 618 else 619 return 0; 620 } 621 622 if (!(snap_seg = find_cow(lv))) 623 return 1; 624 625 old_origin = snap_seg->origin; 626 627 /* Was this the last active snapshot with this origin? */ 628 dm_list_iterate_items(lvl, active_head) { 629 active = lvl->lv; 630 if ((snap_seg = find_cow(active)) && 631 snap_seg->origin == old_origin) { 632 return 1; 633 } 634 } 635 636 #endif 637 638 /*************************/ 639 /* NEW CODE STARTS HERE */ 640 /*************************/ 641 642 int dev_manager_lv_mknodes(const struct logical_volume *lv) 643 { 644 char *name; 645 646 if (!(name = build_dm_name(lv->vg->cmd->mem, lv->vg->name, 647 lv->name, NULL))) 648 return_0; 649 650 return fs_add_lv(lv, name); 651 } 652 653 int dev_manager_lv_rmnodes(const struct logical_volume *lv) 654 { 655 return fs_del_lv(lv); 656 } 657 658 static int _add_dev_to_dtree(struct dev_manager *dm, struct dm_tree *dtree, 659 struct logical_volume *lv, const char *layer) 660 { 661 char *dlid, *name; 662 struct dm_info info, info2; 663 664 if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, layer))) 665 return_0; 666 667 if (!(dlid = build_dlid(dm, lv->lvid.s, layer))) 668 return_0; 669 670 log_debug("Getting device info for %s [%s]", name, dlid); 671 if (!_info(name, dlid, 0, 1, 0, &info, NULL)) { 672 log_error("Failed to get info for %s [%s].", name, dlid); 673 return 0; 674 } 675 676 /* 677 * For top level volumes verify that existing device match 678 * requested major/minor and that major/minor pair is available for use 679 */ 680 if (!layer && lv->major != -1 && lv->minor != -1) { 681 /* 682 * FIXME compare info.major with lv->major if multiple major support 683 */ 684 if (info.exists && (info.minor != lv->minor)) { 685 log_error("Volume %s (%" PRIu32 ":%" PRIu32")" 686 " differs from already active device " 687 "(%" PRIu32 ":%" PRIu32")", 688 lv->name, lv->major, lv->minor, info.major, info.minor); 689 return 0; 690 } 691 if (!info.exists && _info_by_dev(lv->major, lv->minor, &info2) && 692 info2.exists) { 693 log_error("The requested major:minor pair " 694 "(%" PRIu32 ":%" PRIu32") is already used", 695 lv->major, lv->minor); 696 return 0; 697 } 698 } 699 700 if (info.exists && !dm_tree_add_dev(dtree, info.major, info.minor)) { 701 log_error("Failed to add device (%" PRIu32 ":%" PRIu32") to dtree", 702 info.major, info.minor); 703 return 0; 704 } 705 706 return 1; 707 } 708 709 /* 710 * Add LV and any known dependencies 711 */ 712 static int _add_lv_to_dtree(struct dev_manager *dm, struct dm_tree *dtree, struct logical_volume *lv) 713 { 714 if (!_add_dev_to_dtree(dm, dtree, lv, NULL)) 715 return_0; 716 717 /* FIXME Can we avoid doing this every time? */ 718 if (!_add_dev_to_dtree(dm, dtree, lv, "real")) 719 return_0; 720 721 if (!_add_dev_to_dtree(dm, dtree, lv, "cow")) 722 return_0; 723 724 if (!_add_dev_to_dtree(dm, dtree, lv, "_mlog")) 725 return_0; 726 727 return 1; 728 } 729 730 static struct dm_tree *_create_partial_dtree(struct dev_manager *dm, struct logical_volume *lv) 731 { 732 struct dm_tree *dtree; 733 struct dm_list *snh, *snht; 734 struct lv_segment *seg; 735 uint32_t s; 736 737 if (!(dtree = dm_tree_create())) { 738 log_error("Partial dtree creation failed for %s.", lv->name); 739 return NULL; 740 } 741 742 if (!_add_lv_to_dtree(dm, dtree, lv)) 743 goto_bad; 744 745 /* Add any snapshots of this LV */ 746 dm_list_iterate_safe(snh, snht, &lv->snapshot_segs) 747 if (!_add_lv_to_dtree(dm, dtree, dm_list_struct_base(snh, struct lv_segment, origin_list)->cow)) 748 goto_bad; 749 750 /* Add any LVs used by segments in this LV */ 751 dm_list_iterate_items(seg, &lv->segments) 752 for (s = 0; s < seg->area_count; s++) 753 if (seg_type(seg, s) == AREA_LV && seg_lv(seg, s)) { 754 if (!_add_lv_to_dtree(dm, dtree, seg_lv(seg, s))) 755 goto_bad; 756 } 757 758 return dtree; 759 760 bad: 761 dm_tree_free(dtree); 762 return NULL; 763 } 764 765 static char *_add_error_device(struct dev_manager *dm, struct dm_tree *dtree, 766 struct lv_segment *seg, int s) 767 { 768 char *id, *name; 769 char errid[32]; 770 struct dm_tree_node *node; 771 struct lv_segment *seg_i; 772 int segno = -1, i = 0;; 773 uint64_t size = seg->len * seg->lv->vg->extent_size; 774 775 dm_list_iterate_items(seg_i, &seg->lv->segments) { 776 if (seg == seg_i) 777 segno = i; 778 ++i; 779 } 780 781 if (segno < 0) { 782 log_error("_add_error_device called with bad segment"); 783 return_NULL; 784 } 785 786 sprintf(errid, "missing_%d_%d", segno, s); 787 788 if (!(id = build_dlid(dm, seg->lv->lvid.s, errid))) 789 return_NULL; 790 791 if (!(name = build_dm_name(dm->mem, seg->lv->vg->name, 792 seg->lv->name, errid))) 793 return_NULL; 794 if (!(node = dm_tree_add_new_dev(dtree, name, id, 0, 0, 0, 0, 0))) 795 return_NULL; 796 if (!dm_tree_node_add_error_target(node, size)) 797 return_NULL; 798 799 return id; 800 } 801 802 static int _add_error_area(struct dev_manager *dm, struct dm_tree_node *node, 803 struct lv_segment *seg, int s) 804 { 805 char *dlid; 806 uint64_t extent_size = seg->lv->vg->extent_size; 807 808 if (!strcmp(dm->cmd->stripe_filler, "error")) { 809 /* 810 * FIXME, the tree pointer is first field of dm_tree_node, but 811 * we don't have the struct definition available. 812 */ 813 struct dm_tree **tree = (struct dm_tree **) node; 814 dlid = _add_error_device(dm, *tree, seg, s); 815 if (!dlid) 816 return_0; 817 dm_tree_node_add_target_area(node, NULL, dlid, 818 extent_size * seg_le(seg, s)); 819 } else 820 dm_tree_node_add_target_area(node, 821 dm->cmd->stripe_filler, 822 NULL, UINT64_C(0)); 823 824 return 1; 825 } 826 827 int add_areas_line(struct dev_manager *dm, struct lv_segment *seg, 828 struct dm_tree_node *node, uint32_t start_area, 829 uint32_t areas) 830 { 831 uint64_t extent_size = seg->lv->vg->extent_size; 832 uint32_t s; 833 char *dlid; 834 835 for (s = start_area; s < areas; s++) { 836 if ((seg_type(seg, s) == AREA_PV && 837 (!seg_pvseg(seg, s) || 838 !seg_pv(seg, s) || 839 !seg_dev(seg, s))) || 840 (seg_type(seg, s) == AREA_LV && !seg_lv(seg, s))) { 841 if (!_add_error_area(dm, node, seg, s)) 842 return_0; 843 } else if (seg_type(seg, s) == AREA_PV) 844 dm_tree_node_add_target_area(node, 845 dev_name(seg_dev(seg, s)), 846 NULL, 847 (seg_pv(seg, s)->pe_start + 848 (extent_size * seg_pe(seg, s)))); 849 else if (seg_type(seg, s) == AREA_LV) { 850 if (!(dlid = build_dlid(dm, 851 seg_lv(seg, s)->lvid.s, 852 NULL))) 853 return_0; 854 dm_tree_node_add_target_area(node, NULL, dlid, 855 extent_size * seg_le(seg, s)); 856 } else { 857 log_error("Internal error: Unassigned area found in LV %s.", 858 seg->lv->name); 859 return 0; 860 } 861 } 862 863 return 1; 864 } 865 866 static int _add_origin_target_to_dtree(struct dev_manager *dm, 867 struct dm_tree_node *dnode, 868 struct logical_volume *lv) 869 { 870 const char *real_dlid; 871 872 if (!(real_dlid = build_dlid(dm, lv->lvid.s, "real"))) 873 return_0; 874 875 if (!dm_tree_node_add_snapshot_origin_target(dnode, lv->size, real_dlid)) 876 return_0; 877 878 return 1; 879 } 880 881 static int _add_snapshot_target_to_dtree(struct dev_manager *dm, 882 struct dm_tree_node *dnode, 883 struct logical_volume *lv) 884 { 885 const char *origin_dlid; 886 const char *cow_dlid; 887 struct lv_segment *snap_seg; 888 uint64_t size; 889 890 if (!(snap_seg = find_cow(lv))) { 891 log_error("Couldn't find snapshot for '%s'.", lv->name); 892 return 0; 893 } 894 895 if (!(origin_dlid = build_dlid(dm, snap_seg->origin->lvid.s, "real"))) 896 return_0; 897 898 if (!(cow_dlid = build_dlid(dm, snap_seg->cow->lvid.s, "cow"))) 899 return_0; 900 901 size = (uint64_t) snap_seg->len * snap_seg->origin->vg->extent_size; 902 903 if (!dm_tree_node_add_snapshot_target(dnode, size, origin_dlid, cow_dlid, 1, snap_seg->chunk_size)) 904 return_0; 905 906 return 1; 907 } 908 909 static int _add_target_to_dtree(struct dev_manager *dm, 910 struct dm_tree_node *dnode, 911 struct lv_segment *seg) 912 { 913 uint64_t extent_size = seg->lv->vg->extent_size; 914 915 if (!seg->segtype->ops->add_target_line) { 916 log_error("_emit_target: Internal error: Can't handle " 917 "segment type %s", seg->segtype->name); 918 return 0; 919 } 920 921 return seg->segtype->ops->add_target_line(dm, dm->mem, dm->cmd, 922 &dm->target_state, seg, 923 dnode, 924 extent_size * seg->len, 925 &dm-> pvmove_mirror_count); 926 } 927 928 static int _add_new_lv_to_dtree(struct dev_manager *dm, struct dm_tree *dtree, 929 struct logical_volume *lv, const char *layer); 930 931 static int _add_segment_to_dtree(struct dev_manager *dm, 932 struct dm_tree *dtree, 933 struct dm_tree_node *dnode, 934 struct lv_segment *seg, 935 const char *layer) 936 { 937 uint32_t s; 938 struct dm_list *snh; 939 struct lv_segment *seg_present; 940 941 /* Ensure required device-mapper targets are loaded */ 942 seg_present = find_cow(seg->lv) ? : seg; 943 944 log_debug("Checking kernel supports %s segment type for %s%s%s", 945 seg_present->segtype->name, seg->lv->name, 946 layer ? "-" : "", layer ? : ""); 947 948 if (seg_present->segtype->ops->target_present && 949 !seg_present->segtype->ops->target_present(seg_present->lv->vg->cmd, 950 seg_present, NULL)) { 951 log_error("Can't expand LV %s: %s target support missing " 952 "from kernel?", seg->lv->name, seg_present->segtype->name); 953 return 0; 954 } 955 956 /* Add mirror log */ 957 if (seg->log_lv && 958 !_add_new_lv_to_dtree(dm, dtree, seg->log_lv, NULL)) 959 return_0; 960 961 /* If this is a snapshot origin, add real LV */ 962 if (lv_is_origin(seg->lv) && !layer) { 963 if (vg_is_clustered(seg->lv->vg)) { 964 log_error("Clustered snapshots are not yet supported"); 965 return 0; 966 } 967 if (!_add_new_lv_to_dtree(dm, dtree, seg->lv, "real")) 968 return_0; 969 } else if (lv_is_cow(seg->lv) && !layer) { 970 if (!_add_new_lv_to_dtree(dm, dtree, seg->lv, "cow")) 971 return_0; 972 } else { 973 /* Add any LVs used by this segment */ 974 for (s = 0; s < seg->area_count; s++) 975 if ((seg_type(seg, s) == AREA_LV) && 976 (!_add_new_lv_to_dtree(dm, dtree, seg_lv(seg, s), NULL))) 977 return_0; 978 } 979 980 /* Now we've added its dependencies, we can add the target itself */ 981 if (lv_is_origin(seg->lv) && !layer) { 982 if (!_add_origin_target_to_dtree(dm, dnode, seg->lv)) 983 return_0; 984 } else if (lv_is_cow(seg->lv) && !layer) { 985 if (!_add_snapshot_target_to_dtree(dm, dnode, seg->lv)) 986 return_0; 987 } else if (!_add_target_to_dtree(dm, dnode, seg)) 988 return_0; 989 990 if (lv_is_origin(seg->lv) && !layer) 991 /* Add any snapshots of this LV */ 992 dm_list_iterate(snh, &seg->lv->snapshot_segs) 993 if (!_add_new_lv_to_dtree(dm, dtree, dm_list_struct_base(snh, struct lv_segment, origin_list)->cow, NULL)) 994 return_0; 995 996 return 1; 997 } 998 999 static int _add_new_lv_to_dtree(struct dev_manager *dm, struct dm_tree *dtree, 1000 struct logical_volume *lv, const char *layer) 1001 { 1002 struct lv_segment *seg; 1003 struct lv_layer *lvlayer; 1004 struct dm_tree_node *dnode; 1005 char *name, *dlid; 1006 uint32_t max_stripe_size = UINT32_C(0); 1007 uint32_t read_ahead = lv->read_ahead; 1008 uint32_t read_ahead_flags = UINT32_C(0); 1009 uint16_t udev_flags = 0; 1010 1011 if (!(name = build_dm_name(dm->mem, lv->vg->name, lv->name, layer))) 1012 return_0; 1013 1014 if (!(dlid = build_dlid(dm, lv->lvid.s, layer))) 1015 return_0; 1016 1017 /* We've already processed this node if it already has a context ptr */ 1018 if ((dnode = dm_tree_find_node_by_uuid(dtree, dlid)) && 1019 dm_tree_node_get_context(dnode)) 1020 return 1; 1021 1022 if (!(lvlayer = dm_pool_alloc(dm->mem, sizeof(*lvlayer)))) { 1023 log_error("_add_new_lv_to_dtree: pool alloc failed for %s %s.", lv->name, layer); 1024 return 0; 1025 } 1026 1027 lvlayer->lv = lv; 1028 1029 if (layer || !lv_is_visible(lv)) 1030 udev_flags |= DM_UDEV_DISABLE_SUBSYSTEM_RULES_FLAG | 1031 DM_UDEV_DISABLE_DISK_RULES_FLAG | 1032 DM_UDEV_DISABLE_OTHER_RULES_FLAG; 1033 1034 if (lv_is_cow(lv)) 1035 udev_flags |= DM_UDEV_LOW_PRIORITY_FLAG; 1036 1037 /* 1038 * Add LV to dtree. 1039 * If we're working with precommitted metadata, clear any 1040 * existing inactive table left behind. 1041 * Major/minor settings only apply to the visible layer. 1042 */ 1043 if (!(dnode = dm_tree_add_new_dev_with_udev_flags(dtree, name, dlid, 1044 layer ? UINT32_C(0) : (uint32_t) lv->major, 1045 layer ? UINT32_C(0) : (uint32_t) lv->minor, 1046 _read_only_lv(lv), 1047 (lv->vg->status & PRECOMMITTED) ? 1 : 0, 1048 lvlayer, 1049 udev_flags))) 1050 return_0; 1051 1052 /* Store existing name so we can do rename later */ 1053 lvlayer->old_name = dm_tree_node_get_name(dnode); 1054 1055 /* Create table */ 1056 dm->pvmove_mirror_count = 0u; 1057 dm_list_iterate_items(seg, &lv->segments) { 1058 if (!_add_segment_to_dtree(dm, dtree, dnode, seg, layer)) 1059 return_0; 1060 /* These aren't real segments in the LVM2 metadata */ 1061 if (lv_is_origin(lv) && !layer) 1062 break; 1063 if (lv_is_cow(lv) && !layer) 1064 break; 1065 if (max_stripe_size < seg->stripe_size * seg->area_count) 1066 max_stripe_size = seg->stripe_size * seg->area_count; 1067 } 1068 1069 if (read_ahead == DM_READ_AHEAD_AUTO) { 1070 /* we need RA at least twice a whole stripe - see the comment in md/raid0.c */ 1071 read_ahead = max_stripe_size * 2; 1072 if (!read_ahead) 1073 lv_calculate_readahead(lv, &read_ahead); 1074 read_ahead_flags = DM_READ_AHEAD_MINIMUM_FLAG; 1075 } 1076 1077 dm_tree_node_set_read_ahead(dnode, read_ahead, read_ahead_flags); 1078 1079 return 1; 1080 } 1081 1082 /* FIXME: symlinks should be created/destroyed at the same time 1083 * as the kernel devices but we can't do that from within libdevmapper 1084 * at present so we must walk the tree twice instead. */ 1085 1086 /* 1087 * Create LV symlinks for children of supplied root node. 1088 */ 1089 static int _create_lv_symlinks(struct dev_manager *dm, struct dm_tree_node *root) 1090 { 1091 void *handle = NULL; 1092 struct dm_tree_node *child; 1093 struct lv_layer *lvlayer; 1094 char *old_vgname, *old_lvname, *old_layer; 1095 char *new_vgname, *new_lvname, *new_layer; 1096 const char *name; 1097 int r = 1; 1098 1099 while ((child = dm_tree_next_child(&handle, root, 0))) { 1100 if (!(lvlayer = (struct lv_layer *) dm_tree_node_get_context(child))) 1101 continue; 1102 1103 /* Detect rename */ 1104 name = dm_tree_node_get_name(child); 1105 1106 if (name && lvlayer->old_name && *lvlayer->old_name && strcmp(name, lvlayer->old_name)) { 1107 if (!dm_split_lvm_name(dm->mem, lvlayer->old_name, &old_vgname, &old_lvname, &old_layer)) { 1108 log_error("_create_lv_symlinks: Couldn't split up old device name %s", lvlayer->old_name); 1109 return 0; 1110 } 1111 if (!dm_split_lvm_name(dm->mem, name, &new_vgname, &new_lvname, &new_layer)) { 1112 log_error("_create_lv_symlinks: Couldn't split up new device name %s", name); 1113 return 0; 1114 } 1115 if (!fs_rename_lv(lvlayer->lv, name, old_vgname, old_lvname)) 1116 r = 0; 1117 continue; 1118 } 1119 if (lv_is_visible(lvlayer->lv)) { 1120 if (!dev_manager_lv_mknodes(lvlayer->lv)) 1121 r = 0; 1122 continue; 1123 } 1124 if (!dev_manager_lv_rmnodes(lvlayer->lv)) 1125 r = 0; 1126 } 1127 1128 return r; 1129 } 1130 1131 /* 1132 * Remove LV symlinks for children of supplied root node. 1133 */ 1134 static int _remove_lv_symlinks(struct dev_manager *dm, struct dm_tree_node *root) 1135 { 1136 void *handle = NULL; 1137 struct dm_tree_node *child; 1138 char *vgname, *lvname, *layer; 1139 int r = 1; 1140 1141 while ((child = dm_tree_next_child(&handle, root, 0))) { 1142 if (!dm_split_lvm_name(dm->mem, dm_tree_node_get_name(child), &vgname, &lvname, &layer)) { 1143 r = 0; 1144 continue; 1145 } 1146 1147 if (!*vgname) 1148 continue; 1149 1150 /* only top level layer has symlinks */ 1151 if (*layer) 1152 continue; 1153 1154 fs_del_lv_byname(dm->cmd->dev_dir, vgname, lvname); 1155 } 1156 1157 return r; 1158 } 1159 1160 static int _clean_tree(struct dev_manager *dm, struct dm_tree_node *root) 1161 { 1162 void *handle = NULL; 1163 struct dm_tree_node *child; 1164 char *vgname, *lvname, *layer; 1165 const char *name, *uuid; 1166 int r; 1167 1168 while ((child = dm_tree_next_child(&handle, root, 0))) { 1169 if (!(name = dm_tree_node_get_name(child))) 1170 continue; 1171 1172 if (!(uuid = dm_tree_node_get_uuid(child))) 1173 continue; 1174 1175 if (!dm_split_lvm_name(dm->mem, name, &vgname, &lvname, &layer)) { 1176 log_error("_clean_tree: Couldn't split up device name %s.", name); 1177 return 0; 1178 } 1179 1180 /* Not meant to be top level? */ 1181 if (!*layer) 1182 continue; 1183 1184 dm_tree_set_cookie(root, 0); 1185 r = dm_tree_deactivate_children(root, uuid, strlen(uuid)); 1186 if (!dm_udev_wait(dm_tree_get_cookie(root))) 1187 stack; 1188 1189 if (!r) 1190 return_0; 1191 } 1192 1193 return 1; 1194 } 1195 1196 static int _tree_action(struct dev_manager *dm, struct logical_volume *lv, action_t action) 1197 { 1198 struct dm_tree *dtree; 1199 struct dm_tree_node *root; 1200 char *dlid; 1201 int r = 0; 1202 1203 if (!(dtree = _create_partial_dtree(dm, lv))) 1204 return_0; 1205 1206 if (!(root = dm_tree_find_node(dtree, 0, 0))) { 1207 log_error("Lost dependency tree root node"); 1208 goto out; 1209 } 1210 1211 if (!(dlid = build_dlid(dm, lv->lvid.s, NULL))) 1212 goto_out; 1213 1214 /* Only process nodes with uuid of "LVM-" plus VG id. */ 1215 switch(action) { 1216 case CLEAN: 1217 /* Deactivate any unused non-toplevel nodes */ 1218 if (!_clean_tree(dm, root)) 1219 goto_out; 1220 break; 1221 case DEACTIVATE: 1222 /* Deactivate LV and all devices it references that nothing else has open. */ 1223 dm_tree_set_cookie(root, 0); 1224 r = dm_tree_deactivate_children(root, dlid, ID_LEN + sizeof(UUID_PREFIX) - 1); 1225 if (!dm_udev_wait(dm_tree_get_cookie(root))) 1226 stack; 1227 if (!r) 1228 goto_out; 1229 if (!_remove_lv_symlinks(dm, root)) 1230 log_error("Failed to remove all device symlinks associated with %s.", lv->name); 1231 break; 1232 case SUSPEND: 1233 dm_tree_skip_lockfs(root); 1234 if (!dm->flush_required && (lv->status & MIRRORED) && !(lv->status & PVMOVE)) 1235 dm_tree_use_no_flush_suspend(root); 1236 case SUSPEND_WITH_LOCKFS: 1237 if (!dm_tree_suspend_children(root, dlid, ID_LEN + sizeof(UUID_PREFIX) - 1)) 1238 goto_out; 1239 break; 1240 case PRELOAD: 1241 case ACTIVATE: 1242 /* Add all required new devices to tree */ 1243 if (!_add_new_lv_to_dtree(dm, dtree, lv, NULL)) 1244 goto_out; 1245 1246 /* Preload any devices required before any suspensions */ 1247 dm_tree_set_cookie(root, 0); 1248 r = dm_tree_preload_children(root, dlid, ID_LEN + sizeof(UUID_PREFIX) - 1); 1249 if (!dm_udev_wait(dm_tree_get_cookie(root))) 1250 stack; 1251 if (!r) 1252 goto_out; 1253 1254 if (dm_tree_node_size_changed(root)) 1255 dm->flush_required = 1; 1256 1257 if (action == ACTIVATE) { 1258 dm_tree_set_cookie(root, 0); 1259 r = dm_tree_activate_children(root, dlid, ID_LEN + sizeof(UUID_PREFIX) - 1); 1260 if (!dm_udev_wait(dm_tree_get_cookie(root))) 1261 stack; 1262 if (!r) 1263 goto_out; 1264 if (!_create_lv_symlinks(dm, root)) { 1265 log_error("Failed to create symlinks for %s.", lv->name); 1266 goto out; 1267 } 1268 } 1269 1270 break; 1271 default: 1272 log_error("_tree_action: Action %u not supported.", action); 1273 goto out; 1274 } 1275 1276 r = 1; 1277 1278 out: 1279 dm_tree_free(dtree); 1280 1281 return r; 1282 } 1283 1284 int dev_manager_activate(struct dev_manager *dm, struct logical_volume *lv) 1285 { 1286 if (!_tree_action(dm, lv, ACTIVATE)) 1287 return_0; 1288 1289 return _tree_action(dm, lv, CLEAN); 1290 } 1291 1292 int dev_manager_preload(struct dev_manager *dm, struct logical_volume *lv, 1293 int *flush_required) 1294 { 1295 /* FIXME Update the pvmove implementation! */ 1296 if ((lv->status & PVMOVE) || (lv->status & LOCKED)) 1297 return 1; 1298 1299 if (!_tree_action(dm, lv, PRELOAD)) 1300 return 0; 1301 1302 *flush_required = dm->flush_required; 1303 1304 return 1; 1305 } 1306 1307 int dev_manager_deactivate(struct dev_manager *dm, struct logical_volume *lv) 1308 { 1309 int r; 1310 1311 r = _tree_action(dm, lv, DEACTIVATE); 1312 1313 fs_del_lv(lv); 1314 1315 return r; 1316 } 1317 1318 int dev_manager_suspend(struct dev_manager *dm, struct logical_volume *lv, 1319 int lockfs, int flush_required) 1320 { 1321 dm->flush_required = flush_required; 1322 1323 return _tree_action(dm, lv, lockfs ? SUSPEND_WITH_LOCKFS : SUSPEND); 1324 } 1325 1326 /* 1327 * Does device use VG somewhere in its construction? 1328 * Returns 1 if uncertain. 1329 */ 1330 int dev_manager_device_uses_vg(struct device *dev, 1331 struct volume_group *vg) 1332 { 1333 struct dm_tree *dtree; 1334 struct dm_tree_node *root; 1335 char dlid[sizeof(UUID_PREFIX) + sizeof(struct id) - 1] __attribute((aligned(8))); 1336 int r = 1; 1337 1338 if (!(dtree = dm_tree_create())) { 1339 log_error("partial dtree creation failed"); 1340 return r; 1341 } 1342 1343 if (!dm_tree_add_dev(dtree, (uint32_t) MAJOR(dev->dev), (uint32_t) MINOR(dev->dev))) { 1344 log_error("Failed to add device %s (%" PRIu32 ":%" PRIu32") to dtree", 1345 dev_name(dev), (uint32_t) MAJOR(dev->dev), (uint32_t) MINOR(dev->dev)); 1346 goto out; 1347 } 1348 1349 memcpy(dlid, UUID_PREFIX, sizeof(UUID_PREFIX) - 1); 1350 memcpy(dlid + sizeof(UUID_PREFIX) - 1, &vg->id.uuid[0], sizeof(vg->id)); 1351 1352 if (!(root = dm_tree_find_node(dtree, 0, 0))) { 1353 log_error("Lost dependency tree root node"); 1354 goto out; 1355 } 1356 1357 if (dm_tree_children_use_uuid(root, dlid, sizeof(UUID_PREFIX) + sizeof(vg->id) - 1)) 1358 goto_out; 1359 1360 r = 0; 1361 1362 out: 1363 dm_tree_free(dtree); 1364 return r; 1365 } 1366