1 2 #include <linux/ceph/ceph_debug.h> 3 4 #include <linux/module.h> 5 #include <linux/slab.h> 6 #include <asm/div64.h> 7 8 #include <linux/ceph/libceph.h> 9 #include <linux/ceph/osdmap.h> 10 #include <linux/ceph/decode.h> 11 #include <linux/crush/hash.h> 12 #include <linux/crush/mapper.h> 13 14 char *ceph_osdmap_state_str(char *str, int len, int state) 15 { 16 if (!len) 17 return str; 18 19 if ((state & CEPH_OSD_EXISTS) && (state & CEPH_OSD_UP)) 20 snprintf(str, len, "exists, up"); 21 else if (state & CEPH_OSD_EXISTS) 22 snprintf(str, len, "exists"); 23 else if (state & CEPH_OSD_UP) 24 snprintf(str, len, "up"); 25 else 26 snprintf(str, len, "doesn't exist"); 27 28 return str; 29 } 30 31 /* maps */ 32 33 static int calc_bits_of(unsigned int t) 34 { 35 int b = 0; 36 while (t) { 37 t = t >> 1; 38 b++; 39 } 40 return b; 41 } 42 43 /* 44 * the foo_mask is the smallest value 2^n-1 that is >= foo. 45 */ 46 static void calc_pg_masks(struct ceph_pg_pool_info *pi) 47 { 48 pi->pg_num_mask = (1 << calc_bits_of(pi->pg_num-1)) - 1; 49 pi->pgp_num_mask = (1 << calc_bits_of(pi->pgp_num-1)) - 1; 50 } 51 52 /* 53 * decode crush map 54 */ 55 static int crush_decode_uniform_bucket(void **p, void *end, 56 struct crush_bucket_uniform *b) 57 { 58 dout("crush_decode_uniform_bucket %p to %p\n", *p, end); 59 ceph_decode_need(p, end, (1+b->h.size) * sizeof(u32), bad); 60 b->item_weight = ceph_decode_32(p); 61 return 0; 62 bad: 63 return -EINVAL; 64 } 65 66 static int crush_decode_list_bucket(void **p, void *end, 67 struct crush_bucket_list *b) 68 { 69 int j; 70 dout("crush_decode_list_bucket %p to %p\n", *p, end); 71 b->item_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS); 72 if (b->item_weights == NULL) 73 return -ENOMEM; 74 b->sum_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS); 75 if (b->sum_weights == NULL) 76 return -ENOMEM; 77 ceph_decode_need(p, end, 2 * b->h.size * sizeof(u32), bad); 78 for (j = 0; j < b->h.size; j++) { 79 b->item_weights[j] = ceph_decode_32(p); 80 b->sum_weights[j] = ceph_decode_32(p); 81 } 82 return 0; 83 bad: 84 return -EINVAL; 85 } 86 87 static int crush_decode_tree_bucket(void **p, void *end, 88 struct crush_bucket_tree *b) 89 { 90 int j; 91 dout("crush_decode_tree_bucket %p to %p\n", *p, end); 92 ceph_decode_8_safe(p, end, b->num_nodes, bad); 93 b->node_weights = kcalloc(b->num_nodes, sizeof(u32), GFP_NOFS); 94 if (b->node_weights == NULL) 95 return -ENOMEM; 96 ceph_decode_need(p, end, b->num_nodes * sizeof(u32), bad); 97 for (j = 0; j < b->num_nodes; j++) 98 b->node_weights[j] = ceph_decode_32(p); 99 return 0; 100 bad: 101 return -EINVAL; 102 } 103 104 static int crush_decode_straw_bucket(void **p, void *end, 105 struct crush_bucket_straw *b) 106 { 107 int j; 108 dout("crush_decode_straw_bucket %p to %p\n", *p, end); 109 b->item_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS); 110 if (b->item_weights == NULL) 111 return -ENOMEM; 112 b->straws = kcalloc(b->h.size, sizeof(u32), GFP_NOFS); 113 if (b->straws == NULL) 114 return -ENOMEM; 115 ceph_decode_need(p, end, 2 * b->h.size * sizeof(u32), bad); 116 for (j = 0; j < b->h.size; j++) { 117 b->item_weights[j] = ceph_decode_32(p); 118 b->straws[j] = ceph_decode_32(p); 119 } 120 return 0; 121 bad: 122 return -EINVAL; 123 } 124 125 static int crush_decode_straw2_bucket(void **p, void *end, 126 struct crush_bucket_straw2 *b) 127 { 128 int j; 129 dout("crush_decode_straw2_bucket %p to %p\n", *p, end); 130 b->item_weights = kcalloc(b->h.size, sizeof(u32), GFP_NOFS); 131 if (b->item_weights == NULL) 132 return -ENOMEM; 133 ceph_decode_need(p, end, b->h.size * sizeof(u32), bad); 134 for (j = 0; j < b->h.size; j++) 135 b->item_weights[j] = ceph_decode_32(p); 136 return 0; 137 bad: 138 return -EINVAL; 139 } 140 141 static int skip_name_map(void **p, void *end) 142 { 143 int len; 144 ceph_decode_32_safe(p, end, len ,bad); 145 while (len--) { 146 int strlen; 147 *p += sizeof(u32); 148 ceph_decode_32_safe(p, end, strlen, bad); 149 *p += strlen; 150 } 151 return 0; 152 bad: 153 return -EINVAL; 154 } 155 156 static struct crush_map *crush_decode(void *pbyval, void *end) 157 { 158 struct crush_map *c; 159 int err = -EINVAL; 160 int i, j; 161 void **p = &pbyval; 162 void *start = pbyval; 163 u32 magic; 164 u32 num_name_maps; 165 166 dout("crush_decode %p to %p len %d\n", *p, end, (int)(end - *p)); 167 168 c = kzalloc(sizeof(*c), GFP_NOFS); 169 if (c == NULL) 170 return ERR_PTR(-ENOMEM); 171 172 /* set tunables to default values */ 173 c->choose_local_tries = 2; 174 c->choose_local_fallback_tries = 5; 175 c->choose_total_tries = 19; 176 c->chooseleaf_descend_once = 0; 177 178 ceph_decode_need(p, end, 4*sizeof(u32), bad); 179 magic = ceph_decode_32(p); 180 if (magic != CRUSH_MAGIC) { 181 pr_err("crush_decode magic %x != current %x\n", 182 (unsigned int)magic, (unsigned int)CRUSH_MAGIC); 183 goto bad; 184 } 185 c->max_buckets = ceph_decode_32(p); 186 c->max_rules = ceph_decode_32(p); 187 c->max_devices = ceph_decode_32(p); 188 189 c->buckets = kcalloc(c->max_buckets, sizeof(*c->buckets), GFP_NOFS); 190 if (c->buckets == NULL) 191 goto badmem; 192 c->rules = kcalloc(c->max_rules, sizeof(*c->rules), GFP_NOFS); 193 if (c->rules == NULL) 194 goto badmem; 195 196 /* buckets */ 197 for (i = 0; i < c->max_buckets; i++) { 198 int size = 0; 199 u32 alg; 200 struct crush_bucket *b; 201 202 ceph_decode_32_safe(p, end, alg, bad); 203 if (alg == 0) { 204 c->buckets[i] = NULL; 205 continue; 206 } 207 dout("crush_decode bucket %d off %x %p to %p\n", 208 i, (int)(*p-start), *p, end); 209 210 switch (alg) { 211 case CRUSH_BUCKET_UNIFORM: 212 size = sizeof(struct crush_bucket_uniform); 213 break; 214 case CRUSH_BUCKET_LIST: 215 size = sizeof(struct crush_bucket_list); 216 break; 217 case CRUSH_BUCKET_TREE: 218 size = sizeof(struct crush_bucket_tree); 219 break; 220 case CRUSH_BUCKET_STRAW: 221 size = sizeof(struct crush_bucket_straw); 222 break; 223 case CRUSH_BUCKET_STRAW2: 224 size = sizeof(struct crush_bucket_straw2); 225 break; 226 default: 227 err = -EINVAL; 228 goto bad; 229 } 230 BUG_ON(size == 0); 231 b = c->buckets[i] = kzalloc(size, GFP_NOFS); 232 if (b == NULL) 233 goto badmem; 234 235 ceph_decode_need(p, end, 4*sizeof(u32), bad); 236 b->id = ceph_decode_32(p); 237 b->type = ceph_decode_16(p); 238 b->alg = ceph_decode_8(p); 239 b->hash = ceph_decode_8(p); 240 b->weight = ceph_decode_32(p); 241 b->size = ceph_decode_32(p); 242 243 dout("crush_decode bucket size %d off %x %p to %p\n", 244 b->size, (int)(*p-start), *p, end); 245 246 b->items = kcalloc(b->size, sizeof(__s32), GFP_NOFS); 247 if (b->items == NULL) 248 goto badmem; 249 b->perm = kcalloc(b->size, sizeof(u32), GFP_NOFS); 250 if (b->perm == NULL) 251 goto badmem; 252 b->perm_n = 0; 253 254 ceph_decode_need(p, end, b->size*sizeof(u32), bad); 255 for (j = 0; j < b->size; j++) 256 b->items[j] = ceph_decode_32(p); 257 258 switch (b->alg) { 259 case CRUSH_BUCKET_UNIFORM: 260 err = crush_decode_uniform_bucket(p, end, 261 (struct crush_bucket_uniform *)b); 262 if (err < 0) 263 goto bad; 264 break; 265 case CRUSH_BUCKET_LIST: 266 err = crush_decode_list_bucket(p, end, 267 (struct crush_bucket_list *)b); 268 if (err < 0) 269 goto bad; 270 break; 271 case CRUSH_BUCKET_TREE: 272 err = crush_decode_tree_bucket(p, end, 273 (struct crush_bucket_tree *)b); 274 if (err < 0) 275 goto bad; 276 break; 277 case CRUSH_BUCKET_STRAW: 278 err = crush_decode_straw_bucket(p, end, 279 (struct crush_bucket_straw *)b); 280 if (err < 0) 281 goto bad; 282 break; 283 case CRUSH_BUCKET_STRAW2: 284 err = crush_decode_straw2_bucket(p, end, 285 (struct crush_bucket_straw2 *)b); 286 if (err < 0) 287 goto bad; 288 break; 289 } 290 } 291 292 /* rules */ 293 dout("rule vec is %p\n", c->rules); 294 for (i = 0; i < c->max_rules; i++) { 295 u32 yes; 296 struct crush_rule *r; 297 298 ceph_decode_32_safe(p, end, yes, bad); 299 if (!yes) { 300 dout("crush_decode NO rule %d off %x %p to %p\n", 301 i, (int)(*p-start), *p, end); 302 c->rules[i] = NULL; 303 continue; 304 } 305 306 dout("crush_decode rule %d off %x %p to %p\n", 307 i, (int)(*p-start), *p, end); 308 309 /* len */ 310 ceph_decode_32_safe(p, end, yes, bad); 311 #if BITS_PER_LONG == 32 312 err = -EINVAL; 313 if (yes > (ULONG_MAX - sizeof(*r)) 314 / sizeof(struct crush_rule_step)) 315 goto bad; 316 #endif 317 r = c->rules[i] = kmalloc(sizeof(*r) + 318 yes*sizeof(struct crush_rule_step), 319 GFP_NOFS); 320 if (r == NULL) 321 goto badmem; 322 dout(" rule %d is at %p\n", i, r); 323 r->len = yes; 324 ceph_decode_copy_safe(p, end, &r->mask, 4, bad); /* 4 u8's */ 325 ceph_decode_need(p, end, r->len*3*sizeof(u32), bad); 326 for (j = 0; j < r->len; j++) { 327 r->steps[j].op = ceph_decode_32(p); 328 r->steps[j].arg1 = ceph_decode_32(p); 329 r->steps[j].arg2 = ceph_decode_32(p); 330 } 331 } 332 333 /* ignore trailing name maps. */ 334 for (num_name_maps = 0; num_name_maps < 3; num_name_maps++) { 335 err = skip_name_map(p, end); 336 if (err < 0) 337 goto done; 338 } 339 340 /* tunables */ 341 ceph_decode_need(p, end, 3*sizeof(u32), done); 342 c->choose_local_tries = ceph_decode_32(p); 343 c->choose_local_fallback_tries = ceph_decode_32(p); 344 c->choose_total_tries = ceph_decode_32(p); 345 dout("crush decode tunable choose_local_tries = %d\n", 346 c->choose_local_tries); 347 dout("crush decode tunable choose_local_fallback_tries = %d\n", 348 c->choose_local_fallback_tries); 349 dout("crush decode tunable choose_total_tries = %d\n", 350 c->choose_total_tries); 351 352 ceph_decode_need(p, end, sizeof(u32), done); 353 c->chooseleaf_descend_once = ceph_decode_32(p); 354 dout("crush decode tunable chooseleaf_descend_once = %d\n", 355 c->chooseleaf_descend_once); 356 357 ceph_decode_need(p, end, sizeof(u8), done); 358 c->chooseleaf_vary_r = ceph_decode_8(p); 359 dout("crush decode tunable chooseleaf_vary_r = %d\n", 360 c->chooseleaf_vary_r); 361 362 /* skip straw_calc_version, allowed_bucket_algs */ 363 ceph_decode_need(p, end, sizeof(u8) + sizeof(u32), done); 364 *p += sizeof(u8) + sizeof(u32); 365 366 ceph_decode_need(p, end, sizeof(u8), done); 367 c->chooseleaf_stable = ceph_decode_8(p); 368 dout("crush decode tunable chooseleaf_stable = %d\n", 369 c->chooseleaf_stable); 370 371 done: 372 dout("crush_decode success\n"); 373 return c; 374 375 badmem: 376 err = -ENOMEM; 377 bad: 378 dout("crush_decode fail %d\n", err); 379 crush_destroy(c); 380 return ERR_PTR(err); 381 } 382 383 int ceph_pg_compare(const struct ceph_pg *lhs, const struct ceph_pg *rhs) 384 { 385 if (lhs->pool < rhs->pool) 386 return -1; 387 if (lhs->pool > rhs->pool) 388 return 1; 389 if (lhs->seed < rhs->seed) 390 return -1; 391 if (lhs->seed > rhs->seed) 392 return 1; 393 394 return 0; 395 } 396 397 /* 398 * rbtree of pg_mapping for handling pg_temp (explicit mapping of pgid 399 * to a set of osds) and primary_temp (explicit primary setting) 400 */ 401 static int __insert_pg_mapping(struct ceph_pg_mapping *new, 402 struct rb_root *root) 403 { 404 struct rb_node **p = &root->rb_node; 405 struct rb_node *parent = NULL; 406 struct ceph_pg_mapping *pg = NULL; 407 int c; 408 409 dout("__insert_pg_mapping %llx %p\n", *(u64 *)&new->pgid, new); 410 while (*p) { 411 parent = *p; 412 pg = rb_entry(parent, struct ceph_pg_mapping, node); 413 c = ceph_pg_compare(&new->pgid, &pg->pgid); 414 if (c < 0) 415 p = &(*p)->rb_left; 416 else if (c > 0) 417 p = &(*p)->rb_right; 418 else 419 return -EEXIST; 420 } 421 422 rb_link_node(&new->node, parent, p); 423 rb_insert_color(&new->node, root); 424 return 0; 425 } 426 427 static struct ceph_pg_mapping *__lookup_pg_mapping(struct rb_root *root, 428 struct ceph_pg pgid) 429 { 430 struct rb_node *n = root->rb_node; 431 struct ceph_pg_mapping *pg; 432 int c; 433 434 while (n) { 435 pg = rb_entry(n, struct ceph_pg_mapping, node); 436 c = ceph_pg_compare(&pgid, &pg->pgid); 437 if (c < 0) { 438 n = n->rb_left; 439 } else if (c > 0) { 440 n = n->rb_right; 441 } else { 442 dout("__lookup_pg_mapping %lld.%x got %p\n", 443 pgid.pool, pgid.seed, pg); 444 return pg; 445 } 446 } 447 return NULL; 448 } 449 450 static int __remove_pg_mapping(struct rb_root *root, struct ceph_pg pgid) 451 { 452 struct ceph_pg_mapping *pg = __lookup_pg_mapping(root, pgid); 453 454 if (pg) { 455 dout("__remove_pg_mapping %lld.%x %p\n", pgid.pool, pgid.seed, 456 pg); 457 rb_erase(&pg->node, root); 458 kfree(pg); 459 return 0; 460 } 461 dout("__remove_pg_mapping %lld.%x dne\n", pgid.pool, pgid.seed); 462 return -ENOENT; 463 } 464 465 /* 466 * rbtree of pg pool info 467 */ 468 static int __insert_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *new) 469 { 470 struct rb_node **p = &root->rb_node; 471 struct rb_node *parent = NULL; 472 struct ceph_pg_pool_info *pi = NULL; 473 474 while (*p) { 475 parent = *p; 476 pi = rb_entry(parent, struct ceph_pg_pool_info, node); 477 if (new->id < pi->id) 478 p = &(*p)->rb_left; 479 else if (new->id > pi->id) 480 p = &(*p)->rb_right; 481 else 482 return -EEXIST; 483 } 484 485 rb_link_node(&new->node, parent, p); 486 rb_insert_color(&new->node, root); 487 return 0; 488 } 489 490 static struct ceph_pg_pool_info *__lookup_pg_pool(struct rb_root *root, u64 id) 491 { 492 struct ceph_pg_pool_info *pi; 493 struct rb_node *n = root->rb_node; 494 495 while (n) { 496 pi = rb_entry(n, struct ceph_pg_pool_info, node); 497 if (id < pi->id) 498 n = n->rb_left; 499 else if (id > pi->id) 500 n = n->rb_right; 501 else 502 return pi; 503 } 504 return NULL; 505 } 506 507 struct ceph_pg_pool_info *ceph_pg_pool_by_id(struct ceph_osdmap *map, u64 id) 508 { 509 return __lookup_pg_pool(&map->pg_pools, id); 510 } 511 512 const char *ceph_pg_pool_name_by_id(struct ceph_osdmap *map, u64 id) 513 { 514 struct ceph_pg_pool_info *pi; 515 516 if (id == CEPH_NOPOOL) 517 return NULL; 518 519 if (WARN_ON_ONCE(id > (u64) INT_MAX)) 520 return NULL; 521 522 pi = __lookup_pg_pool(&map->pg_pools, (int) id); 523 524 return pi ? pi->name : NULL; 525 } 526 EXPORT_SYMBOL(ceph_pg_pool_name_by_id); 527 528 int ceph_pg_poolid_by_name(struct ceph_osdmap *map, const char *name) 529 { 530 struct rb_node *rbp; 531 532 for (rbp = rb_first(&map->pg_pools); rbp; rbp = rb_next(rbp)) { 533 struct ceph_pg_pool_info *pi = 534 rb_entry(rbp, struct ceph_pg_pool_info, node); 535 if (pi->name && strcmp(pi->name, name) == 0) 536 return pi->id; 537 } 538 return -ENOENT; 539 } 540 EXPORT_SYMBOL(ceph_pg_poolid_by_name); 541 542 static void __remove_pg_pool(struct rb_root *root, struct ceph_pg_pool_info *pi) 543 { 544 rb_erase(&pi->node, root); 545 kfree(pi->name); 546 kfree(pi); 547 } 548 549 static int decode_pool(void **p, void *end, struct ceph_pg_pool_info *pi) 550 { 551 u8 ev, cv; 552 unsigned len, num; 553 void *pool_end; 554 555 ceph_decode_need(p, end, 2 + 4, bad); 556 ev = ceph_decode_8(p); /* encoding version */ 557 cv = ceph_decode_8(p); /* compat version */ 558 if (ev < 5) { 559 pr_warn("got v %d < 5 cv %d of ceph_pg_pool\n", ev, cv); 560 return -EINVAL; 561 } 562 if (cv > 9) { 563 pr_warn("got v %d cv %d > 9 of ceph_pg_pool\n", ev, cv); 564 return -EINVAL; 565 } 566 len = ceph_decode_32(p); 567 ceph_decode_need(p, end, len, bad); 568 pool_end = *p + len; 569 570 pi->type = ceph_decode_8(p); 571 pi->size = ceph_decode_8(p); 572 pi->crush_ruleset = ceph_decode_8(p); 573 pi->object_hash = ceph_decode_8(p); 574 575 pi->pg_num = ceph_decode_32(p); 576 pi->pgp_num = ceph_decode_32(p); 577 578 *p += 4 + 4; /* skip lpg* */ 579 *p += 4; /* skip last_change */ 580 *p += 8 + 4; /* skip snap_seq, snap_epoch */ 581 582 /* skip snaps */ 583 num = ceph_decode_32(p); 584 while (num--) { 585 *p += 8; /* snapid key */ 586 *p += 1 + 1; /* versions */ 587 len = ceph_decode_32(p); 588 *p += len; 589 } 590 591 /* skip removed_snaps */ 592 num = ceph_decode_32(p); 593 *p += num * (8 + 8); 594 595 *p += 8; /* skip auid */ 596 pi->flags = ceph_decode_64(p); 597 *p += 4; /* skip crash_replay_interval */ 598 599 if (ev >= 7) 600 pi->min_size = ceph_decode_8(p); 601 else 602 pi->min_size = pi->size - pi->size / 2; 603 604 if (ev >= 8) 605 *p += 8 + 8; /* skip quota_max_* */ 606 607 if (ev >= 9) { 608 /* skip tiers */ 609 num = ceph_decode_32(p); 610 *p += num * 8; 611 612 *p += 8; /* skip tier_of */ 613 *p += 1; /* skip cache_mode */ 614 615 pi->read_tier = ceph_decode_64(p); 616 pi->write_tier = ceph_decode_64(p); 617 } else { 618 pi->read_tier = -1; 619 pi->write_tier = -1; 620 } 621 622 if (ev >= 10) { 623 /* skip properties */ 624 num = ceph_decode_32(p); 625 while (num--) { 626 len = ceph_decode_32(p); 627 *p += len; /* key */ 628 len = ceph_decode_32(p); 629 *p += len; /* val */ 630 } 631 } 632 633 if (ev >= 11) { 634 /* skip hit_set_params */ 635 *p += 1 + 1; /* versions */ 636 len = ceph_decode_32(p); 637 *p += len; 638 639 *p += 4; /* skip hit_set_period */ 640 *p += 4; /* skip hit_set_count */ 641 } 642 643 if (ev >= 12) 644 *p += 4; /* skip stripe_width */ 645 646 if (ev >= 13) { 647 *p += 8; /* skip target_max_bytes */ 648 *p += 8; /* skip target_max_objects */ 649 *p += 4; /* skip cache_target_dirty_ratio_micro */ 650 *p += 4; /* skip cache_target_full_ratio_micro */ 651 *p += 4; /* skip cache_min_flush_age */ 652 *p += 4; /* skip cache_min_evict_age */ 653 } 654 655 if (ev >= 14) { 656 /* skip erasure_code_profile */ 657 len = ceph_decode_32(p); 658 *p += len; 659 } 660 661 if (ev >= 15) 662 pi->last_force_request_resend = ceph_decode_32(p); 663 else 664 pi->last_force_request_resend = 0; 665 666 /* ignore the rest */ 667 668 *p = pool_end; 669 calc_pg_masks(pi); 670 return 0; 671 672 bad: 673 return -EINVAL; 674 } 675 676 static int decode_pool_names(void **p, void *end, struct ceph_osdmap *map) 677 { 678 struct ceph_pg_pool_info *pi; 679 u32 num, len; 680 u64 pool; 681 682 ceph_decode_32_safe(p, end, num, bad); 683 dout(" %d pool names\n", num); 684 while (num--) { 685 ceph_decode_64_safe(p, end, pool, bad); 686 ceph_decode_32_safe(p, end, len, bad); 687 dout(" pool %llu len %d\n", pool, len); 688 ceph_decode_need(p, end, len, bad); 689 pi = __lookup_pg_pool(&map->pg_pools, pool); 690 if (pi) { 691 char *name = kstrndup(*p, len, GFP_NOFS); 692 693 if (!name) 694 return -ENOMEM; 695 kfree(pi->name); 696 pi->name = name; 697 dout(" name is %s\n", pi->name); 698 } 699 *p += len; 700 } 701 return 0; 702 703 bad: 704 return -EINVAL; 705 } 706 707 /* 708 * osd map 709 */ 710 struct ceph_osdmap *ceph_osdmap_alloc(void) 711 { 712 struct ceph_osdmap *map; 713 714 map = kzalloc(sizeof(*map), GFP_NOIO); 715 if (!map) 716 return NULL; 717 718 map->pg_pools = RB_ROOT; 719 map->pool_max = -1; 720 map->pg_temp = RB_ROOT; 721 map->primary_temp = RB_ROOT; 722 mutex_init(&map->crush_scratch_mutex); 723 724 return map; 725 } 726 727 void ceph_osdmap_destroy(struct ceph_osdmap *map) 728 { 729 dout("osdmap_destroy %p\n", map); 730 if (map->crush) 731 crush_destroy(map->crush); 732 while (!RB_EMPTY_ROOT(&map->pg_temp)) { 733 struct ceph_pg_mapping *pg = 734 rb_entry(rb_first(&map->pg_temp), 735 struct ceph_pg_mapping, node); 736 rb_erase(&pg->node, &map->pg_temp); 737 kfree(pg); 738 } 739 while (!RB_EMPTY_ROOT(&map->primary_temp)) { 740 struct ceph_pg_mapping *pg = 741 rb_entry(rb_first(&map->primary_temp), 742 struct ceph_pg_mapping, node); 743 rb_erase(&pg->node, &map->primary_temp); 744 kfree(pg); 745 } 746 while (!RB_EMPTY_ROOT(&map->pg_pools)) { 747 struct ceph_pg_pool_info *pi = 748 rb_entry(rb_first(&map->pg_pools), 749 struct ceph_pg_pool_info, node); 750 __remove_pg_pool(&map->pg_pools, pi); 751 } 752 kfree(map->osd_state); 753 kfree(map->osd_weight); 754 kfree(map->osd_addr); 755 kfree(map->osd_primary_affinity); 756 kfree(map); 757 } 758 759 /* 760 * Adjust max_osd value, (re)allocate arrays. 761 * 762 * The new elements are properly initialized. 763 */ 764 static int osdmap_set_max_osd(struct ceph_osdmap *map, int max) 765 { 766 u8 *state; 767 u32 *weight; 768 struct ceph_entity_addr *addr; 769 int i; 770 771 state = krealloc(map->osd_state, max*sizeof(*state), GFP_NOFS); 772 if (!state) 773 return -ENOMEM; 774 map->osd_state = state; 775 776 weight = krealloc(map->osd_weight, max*sizeof(*weight), GFP_NOFS); 777 if (!weight) 778 return -ENOMEM; 779 map->osd_weight = weight; 780 781 addr = krealloc(map->osd_addr, max*sizeof(*addr), GFP_NOFS); 782 if (!addr) 783 return -ENOMEM; 784 map->osd_addr = addr; 785 786 for (i = map->max_osd; i < max; i++) { 787 map->osd_state[i] = 0; 788 map->osd_weight[i] = CEPH_OSD_OUT; 789 memset(map->osd_addr + i, 0, sizeof(*map->osd_addr)); 790 } 791 792 if (map->osd_primary_affinity) { 793 u32 *affinity; 794 795 affinity = krealloc(map->osd_primary_affinity, 796 max*sizeof(*affinity), GFP_NOFS); 797 if (!affinity) 798 return -ENOMEM; 799 map->osd_primary_affinity = affinity; 800 801 for (i = map->max_osd; i < max; i++) 802 map->osd_primary_affinity[i] = 803 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY; 804 } 805 806 map->max_osd = max; 807 808 return 0; 809 } 810 811 #define OSDMAP_WRAPPER_COMPAT_VER 7 812 #define OSDMAP_CLIENT_DATA_COMPAT_VER 1 813 814 /* 815 * Return 0 or error. On success, *v is set to 0 for old (v6) osdmaps, 816 * to struct_v of the client_data section for new (v7 and above) 817 * osdmaps. 818 */ 819 static int get_osdmap_client_data_v(void **p, void *end, 820 const char *prefix, u8 *v) 821 { 822 u8 struct_v; 823 824 ceph_decode_8_safe(p, end, struct_v, e_inval); 825 if (struct_v >= 7) { 826 u8 struct_compat; 827 828 ceph_decode_8_safe(p, end, struct_compat, e_inval); 829 if (struct_compat > OSDMAP_WRAPPER_COMPAT_VER) { 830 pr_warn("got v %d cv %d > %d of %s ceph_osdmap\n", 831 struct_v, struct_compat, 832 OSDMAP_WRAPPER_COMPAT_VER, prefix); 833 return -EINVAL; 834 } 835 *p += 4; /* ignore wrapper struct_len */ 836 837 ceph_decode_8_safe(p, end, struct_v, e_inval); 838 ceph_decode_8_safe(p, end, struct_compat, e_inval); 839 if (struct_compat > OSDMAP_CLIENT_DATA_COMPAT_VER) { 840 pr_warn("got v %d cv %d > %d of %s ceph_osdmap client data\n", 841 struct_v, struct_compat, 842 OSDMAP_CLIENT_DATA_COMPAT_VER, prefix); 843 return -EINVAL; 844 } 845 *p += 4; /* ignore client data struct_len */ 846 } else { 847 u16 version; 848 849 *p -= 1; 850 ceph_decode_16_safe(p, end, version, e_inval); 851 if (version < 6) { 852 pr_warn("got v %d < 6 of %s ceph_osdmap\n", 853 version, prefix); 854 return -EINVAL; 855 } 856 857 /* old osdmap enconding */ 858 struct_v = 0; 859 } 860 861 *v = struct_v; 862 return 0; 863 864 e_inval: 865 return -EINVAL; 866 } 867 868 static int __decode_pools(void **p, void *end, struct ceph_osdmap *map, 869 bool incremental) 870 { 871 u32 n; 872 873 ceph_decode_32_safe(p, end, n, e_inval); 874 while (n--) { 875 struct ceph_pg_pool_info *pi; 876 u64 pool; 877 int ret; 878 879 ceph_decode_64_safe(p, end, pool, e_inval); 880 881 pi = __lookup_pg_pool(&map->pg_pools, pool); 882 if (!incremental || !pi) { 883 pi = kzalloc(sizeof(*pi), GFP_NOFS); 884 if (!pi) 885 return -ENOMEM; 886 887 pi->id = pool; 888 889 ret = __insert_pg_pool(&map->pg_pools, pi); 890 if (ret) { 891 kfree(pi); 892 return ret; 893 } 894 } 895 896 ret = decode_pool(p, end, pi); 897 if (ret) 898 return ret; 899 } 900 901 return 0; 902 903 e_inval: 904 return -EINVAL; 905 } 906 907 static int decode_pools(void **p, void *end, struct ceph_osdmap *map) 908 { 909 return __decode_pools(p, end, map, false); 910 } 911 912 static int decode_new_pools(void **p, void *end, struct ceph_osdmap *map) 913 { 914 return __decode_pools(p, end, map, true); 915 } 916 917 static int __decode_pg_temp(void **p, void *end, struct ceph_osdmap *map, 918 bool incremental) 919 { 920 u32 n; 921 922 ceph_decode_32_safe(p, end, n, e_inval); 923 while (n--) { 924 struct ceph_pg pgid; 925 u32 len, i; 926 int ret; 927 928 ret = ceph_decode_pgid(p, end, &pgid); 929 if (ret) 930 return ret; 931 932 ceph_decode_32_safe(p, end, len, e_inval); 933 934 ret = __remove_pg_mapping(&map->pg_temp, pgid); 935 BUG_ON(!incremental && ret != -ENOENT); 936 937 if (!incremental || len > 0) { 938 struct ceph_pg_mapping *pg; 939 940 ceph_decode_need(p, end, len*sizeof(u32), e_inval); 941 942 if (len > (UINT_MAX - sizeof(*pg)) / sizeof(u32)) 943 return -EINVAL; 944 945 pg = kzalloc(sizeof(*pg) + len*sizeof(u32), GFP_NOFS); 946 if (!pg) 947 return -ENOMEM; 948 949 pg->pgid = pgid; 950 pg->pg_temp.len = len; 951 for (i = 0; i < len; i++) 952 pg->pg_temp.osds[i] = ceph_decode_32(p); 953 954 ret = __insert_pg_mapping(pg, &map->pg_temp); 955 if (ret) { 956 kfree(pg); 957 return ret; 958 } 959 } 960 } 961 962 return 0; 963 964 e_inval: 965 return -EINVAL; 966 } 967 968 static int decode_pg_temp(void **p, void *end, struct ceph_osdmap *map) 969 { 970 return __decode_pg_temp(p, end, map, false); 971 } 972 973 static int decode_new_pg_temp(void **p, void *end, struct ceph_osdmap *map) 974 { 975 return __decode_pg_temp(p, end, map, true); 976 } 977 978 static int __decode_primary_temp(void **p, void *end, struct ceph_osdmap *map, 979 bool incremental) 980 { 981 u32 n; 982 983 ceph_decode_32_safe(p, end, n, e_inval); 984 while (n--) { 985 struct ceph_pg pgid; 986 u32 osd; 987 int ret; 988 989 ret = ceph_decode_pgid(p, end, &pgid); 990 if (ret) 991 return ret; 992 993 ceph_decode_32_safe(p, end, osd, e_inval); 994 995 ret = __remove_pg_mapping(&map->primary_temp, pgid); 996 BUG_ON(!incremental && ret != -ENOENT); 997 998 if (!incremental || osd != (u32)-1) { 999 struct ceph_pg_mapping *pg; 1000 1001 pg = kzalloc(sizeof(*pg), GFP_NOFS); 1002 if (!pg) 1003 return -ENOMEM; 1004 1005 pg->pgid = pgid; 1006 pg->primary_temp.osd = osd; 1007 1008 ret = __insert_pg_mapping(pg, &map->primary_temp); 1009 if (ret) { 1010 kfree(pg); 1011 return ret; 1012 } 1013 } 1014 } 1015 1016 return 0; 1017 1018 e_inval: 1019 return -EINVAL; 1020 } 1021 1022 static int decode_primary_temp(void **p, void *end, struct ceph_osdmap *map) 1023 { 1024 return __decode_primary_temp(p, end, map, false); 1025 } 1026 1027 static int decode_new_primary_temp(void **p, void *end, 1028 struct ceph_osdmap *map) 1029 { 1030 return __decode_primary_temp(p, end, map, true); 1031 } 1032 1033 u32 ceph_get_primary_affinity(struct ceph_osdmap *map, int osd) 1034 { 1035 BUG_ON(osd >= map->max_osd); 1036 1037 if (!map->osd_primary_affinity) 1038 return CEPH_OSD_DEFAULT_PRIMARY_AFFINITY; 1039 1040 return map->osd_primary_affinity[osd]; 1041 } 1042 1043 static int set_primary_affinity(struct ceph_osdmap *map, int osd, u32 aff) 1044 { 1045 BUG_ON(osd >= map->max_osd); 1046 1047 if (!map->osd_primary_affinity) { 1048 int i; 1049 1050 map->osd_primary_affinity = kmalloc(map->max_osd*sizeof(u32), 1051 GFP_NOFS); 1052 if (!map->osd_primary_affinity) 1053 return -ENOMEM; 1054 1055 for (i = 0; i < map->max_osd; i++) 1056 map->osd_primary_affinity[i] = 1057 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY; 1058 } 1059 1060 map->osd_primary_affinity[osd] = aff; 1061 1062 return 0; 1063 } 1064 1065 static int decode_primary_affinity(void **p, void *end, 1066 struct ceph_osdmap *map) 1067 { 1068 u32 len, i; 1069 1070 ceph_decode_32_safe(p, end, len, e_inval); 1071 if (len == 0) { 1072 kfree(map->osd_primary_affinity); 1073 map->osd_primary_affinity = NULL; 1074 return 0; 1075 } 1076 if (len != map->max_osd) 1077 goto e_inval; 1078 1079 ceph_decode_need(p, end, map->max_osd*sizeof(u32), e_inval); 1080 1081 for (i = 0; i < map->max_osd; i++) { 1082 int ret; 1083 1084 ret = set_primary_affinity(map, i, ceph_decode_32(p)); 1085 if (ret) 1086 return ret; 1087 } 1088 1089 return 0; 1090 1091 e_inval: 1092 return -EINVAL; 1093 } 1094 1095 static int decode_new_primary_affinity(void **p, void *end, 1096 struct ceph_osdmap *map) 1097 { 1098 u32 n; 1099 1100 ceph_decode_32_safe(p, end, n, e_inval); 1101 while (n--) { 1102 u32 osd, aff; 1103 int ret; 1104 1105 ceph_decode_32_safe(p, end, osd, e_inval); 1106 ceph_decode_32_safe(p, end, aff, e_inval); 1107 1108 ret = set_primary_affinity(map, osd, aff); 1109 if (ret) 1110 return ret; 1111 1112 pr_info("osd%d primary-affinity 0x%x\n", osd, aff); 1113 } 1114 1115 return 0; 1116 1117 e_inval: 1118 return -EINVAL; 1119 } 1120 1121 /* 1122 * decode a full map. 1123 */ 1124 static int osdmap_decode(void **p, void *end, struct ceph_osdmap *map) 1125 { 1126 u8 struct_v; 1127 u32 epoch = 0; 1128 void *start = *p; 1129 u32 max; 1130 u32 len, i; 1131 int err; 1132 1133 dout("%s %p to %p len %d\n", __func__, *p, end, (int)(end - *p)); 1134 1135 err = get_osdmap_client_data_v(p, end, "full", &struct_v); 1136 if (err) 1137 goto bad; 1138 1139 /* fsid, epoch, created, modified */ 1140 ceph_decode_need(p, end, sizeof(map->fsid) + sizeof(u32) + 1141 sizeof(map->created) + sizeof(map->modified), e_inval); 1142 ceph_decode_copy(p, &map->fsid, sizeof(map->fsid)); 1143 epoch = map->epoch = ceph_decode_32(p); 1144 ceph_decode_copy(p, &map->created, sizeof(map->created)); 1145 ceph_decode_copy(p, &map->modified, sizeof(map->modified)); 1146 1147 /* pools */ 1148 err = decode_pools(p, end, map); 1149 if (err) 1150 goto bad; 1151 1152 /* pool_name */ 1153 err = decode_pool_names(p, end, map); 1154 if (err) 1155 goto bad; 1156 1157 ceph_decode_32_safe(p, end, map->pool_max, e_inval); 1158 1159 ceph_decode_32_safe(p, end, map->flags, e_inval); 1160 1161 /* max_osd */ 1162 ceph_decode_32_safe(p, end, max, e_inval); 1163 1164 /* (re)alloc osd arrays */ 1165 err = osdmap_set_max_osd(map, max); 1166 if (err) 1167 goto bad; 1168 1169 /* osd_state, osd_weight, osd_addrs->client_addr */ 1170 ceph_decode_need(p, end, 3*sizeof(u32) + 1171 map->max_osd*(1 + sizeof(*map->osd_weight) + 1172 sizeof(*map->osd_addr)), e_inval); 1173 1174 if (ceph_decode_32(p) != map->max_osd) 1175 goto e_inval; 1176 1177 ceph_decode_copy(p, map->osd_state, map->max_osd); 1178 1179 if (ceph_decode_32(p) != map->max_osd) 1180 goto e_inval; 1181 1182 for (i = 0; i < map->max_osd; i++) 1183 map->osd_weight[i] = ceph_decode_32(p); 1184 1185 if (ceph_decode_32(p) != map->max_osd) 1186 goto e_inval; 1187 1188 ceph_decode_copy(p, map->osd_addr, map->max_osd*sizeof(*map->osd_addr)); 1189 for (i = 0; i < map->max_osd; i++) 1190 ceph_decode_addr(&map->osd_addr[i]); 1191 1192 /* pg_temp */ 1193 err = decode_pg_temp(p, end, map); 1194 if (err) 1195 goto bad; 1196 1197 /* primary_temp */ 1198 if (struct_v >= 1) { 1199 err = decode_primary_temp(p, end, map); 1200 if (err) 1201 goto bad; 1202 } 1203 1204 /* primary_affinity */ 1205 if (struct_v >= 2) { 1206 err = decode_primary_affinity(p, end, map); 1207 if (err) 1208 goto bad; 1209 } else { 1210 /* XXX can this happen? */ 1211 kfree(map->osd_primary_affinity); 1212 map->osd_primary_affinity = NULL; 1213 } 1214 1215 /* crush */ 1216 ceph_decode_32_safe(p, end, len, e_inval); 1217 map->crush = crush_decode(*p, min(*p + len, end)); 1218 if (IS_ERR(map->crush)) { 1219 err = PTR_ERR(map->crush); 1220 map->crush = NULL; 1221 goto bad; 1222 } 1223 *p += len; 1224 1225 /* ignore the rest */ 1226 *p = end; 1227 1228 dout("full osdmap epoch %d max_osd %d\n", map->epoch, map->max_osd); 1229 return 0; 1230 1231 e_inval: 1232 err = -EINVAL; 1233 bad: 1234 pr_err("corrupt full osdmap (%d) epoch %d off %d (%p of %p-%p)\n", 1235 err, epoch, (int)(*p - start), *p, start, end); 1236 print_hex_dump(KERN_DEBUG, "osdmap: ", 1237 DUMP_PREFIX_OFFSET, 16, 1, 1238 start, end - start, true); 1239 return err; 1240 } 1241 1242 /* 1243 * Allocate and decode a full map. 1244 */ 1245 struct ceph_osdmap *ceph_osdmap_decode(void **p, void *end) 1246 { 1247 struct ceph_osdmap *map; 1248 int ret; 1249 1250 map = ceph_osdmap_alloc(); 1251 if (!map) 1252 return ERR_PTR(-ENOMEM); 1253 1254 ret = osdmap_decode(p, end, map); 1255 if (ret) { 1256 ceph_osdmap_destroy(map); 1257 return ERR_PTR(ret); 1258 } 1259 1260 return map; 1261 } 1262 1263 /* 1264 * Encoding order is (new_up_client, new_state, new_weight). Need to 1265 * apply in the (new_weight, new_state, new_up_client) order, because 1266 * an incremental map may look like e.g. 1267 * 1268 * new_up_client: { osd=6, addr=... } # set osd_state and addr 1269 * new_state: { osd=6, xorstate=EXISTS } # clear osd_state 1270 */ 1271 static int decode_new_up_state_weight(void **p, void *end, 1272 struct ceph_osdmap *map) 1273 { 1274 void *new_up_client; 1275 void *new_state; 1276 void *new_weight_end; 1277 u32 len; 1278 1279 new_up_client = *p; 1280 ceph_decode_32_safe(p, end, len, e_inval); 1281 len *= sizeof(u32) + sizeof(struct ceph_entity_addr); 1282 ceph_decode_need(p, end, len, e_inval); 1283 *p += len; 1284 1285 new_state = *p; 1286 ceph_decode_32_safe(p, end, len, e_inval); 1287 len *= sizeof(u32) + sizeof(u8); 1288 ceph_decode_need(p, end, len, e_inval); 1289 *p += len; 1290 1291 /* new_weight */ 1292 ceph_decode_32_safe(p, end, len, e_inval); 1293 while (len--) { 1294 s32 osd; 1295 u32 w; 1296 1297 ceph_decode_need(p, end, 2*sizeof(u32), e_inval); 1298 osd = ceph_decode_32(p); 1299 w = ceph_decode_32(p); 1300 BUG_ON(osd >= map->max_osd); 1301 pr_info("osd%d weight 0x%x %s\n", osd, w, 1302 w == CEPH_OSD_IN ? "(in)" : 1303 (w == CEPH_OSD_OUT ? "(out)" : "")); 1304 map->osd_weight[osd] = w; 1305 1306 /* 1307 * If we are marking in, set the EXISTS, and clear the 1308 * AUTOOUT and NEW bits. 1309 */ 1310 if (w) { 1311 map->osd_state[osd] |= CEPH_OSD_EXISTS; 1312 map->osd_state[osd] &= ~(CEPH_OSD_AUTOOUT | 1313 CEPH_OSD_NEW); 1314 } 1315 } 1316 new_weight_end = *p; 1317 1318 /* new_state (up/down) */ 1319 *p = new_state; 1320 len = ceph_decode_32(p); 1321 while (len--) { 1322 s32 osd; 1323 u8 xorstate; 1324 int ret; 1325 1326 osd = ceph_decode_32(p); 1327 xorstate = ceph_decode_8(p); 1328 if (xorstate == 0) 1329 xorstate = CEPH_OSD_UP; 1330 BUG_ON(osd >= map->max_osd); 1331 if ((map->osd_state[osd] & CEPH_OSD_UP) && 1332 (xorstate & CEPH_OSD_UP)) 1333 pr_info("osd%d down\n", osd); 1334 if ((map->osd_state[osd] & CEPH_OSD_EXISTS) && 1335 (xorstate & CEPH_OSD_EXISTS)) { 1336 pr_info("osd%d does not exist\n", osd); 1337 map->osd_weight[osd] = CEPH_OSD_IN; 1338 ret = set_primary_affinity(map, osd, 1339 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY); 1340 if (ret) 1341 return ret; 1342 memset(map->osd_addr + osd, 0, sizeof(*map->osd_addr)); 1343 map->osd_state[osd] = 0; 1344 } else { 1345 map->osd_state[osd] ^= xorstate; 1346 } 1347 } 1348 1349 /* new_up_client */ 1350 *p = new_up_client; 1351 len = ceph_decode_32(p); 1352 while (len--) { 1353 s32 osd; 1354 struct ceph_entity_addr addr; 1355 1356 osd = ceph_decode_32(p); 1357 ceph_decode_copy(p, &addr, sizeof(addr)); 1358 ceph_decode_addr(&addr); 1359 BUG_ON(osd >= map->max_osd); 1360 pr_info("osd%d up\n", osd); 1361 map->osd_state[osd] |= CEPH_OSD_EXISTS | CEPH_OSD_UP; 1362 map->osd_addr[osd] = addr; 1363 } 1364 1365 *p = new_weight_end; 1366 return 0; 1367 1368 e_inval: 1369 return -EINVAL; 1370 } 1371 1372 /* 1373 * decode and apply an incremental map update. 1374 */ 1375 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end, 1376 struct ceph_osdmap *map) 1377 { 1378 struct crush_map *newcrush = NULL; 1379 struct ceph_fsid fsid; 1380 u32 epoch = 0; 1381 struct ceph_timespec modified; 1382 s32 len; 1383 u64 pool; 1384 __s64 new_pool_max; 1385 __s32 new_flags, max; 1386 void *start = *p; 1387 int err; 1388 u8 struct_v; 1389 1390 dout("%s %p to %p len %d\n", __func__, *p, end, (int)(end - *p)); 1391 1392 err = get_osdmap_client_data_v(p, end, "inc", &struct_v); 1393 if (err) 1394 goto bad; 1395 1396 /* fsid, epoch, modified, new_pool_max, new_flags */ 1397 ceph_decode_need(p, end, sizeof(fsid) + sizeof(u32) + sizeof(modified) + 1398 sizeof(u64) + sizeof(u32), e_inval); 1399 ceph_decode_copy(p, &fsid, sizeof(fsid)); 1400 epoch = ceph_decode_32(p); 1401 BUG_ON(epoch != map->epoch+1); 1402 ceph_decode_copy(p, &modified, sizeof(modified)); 1403 new_pool_max = ceph_decode_64(p); 1404 new_flags = ceph_decode_32(p); 1405 1406 /* full map? */ 1407 ceph_decode_32_safe(p, end, len, e_inval); 1408 if (len > 0) { 1409 dout("apply_incremental full map len %d, %p to %p\n", 1410 len, *p, end); 1411 return ceph_osdmap_decode(p, min(*p+len, end)); 1412 } 1413 1414 /* new crush? */ 1415 ceph_decode_32_safe(p, end, len, e_inval); 1416 if (len > 0) { 1417 newcrush = crush_decode(*p, min(*p+len, end)); 1418 if (IS_ERR(newcrush)) { 1419 err = PTR_ERR(newcrush); 1420 newcrush = NULL; 1421 goto bad; 1422 } 1423 *p += len; 1424 } 1425 1426 /* new flags? */ 1427 if (new_flags >= 0) 1428 map->flags = new_flags; 1429 if (new_pool_max >= 0) 1430 map->pool_max = new_pool_max; 1431 1432 /* new max? */ 1433 ceph_decode_32_safe(p, end, max, e_inval); 1434 if (max >= 0) { 1435 err = osdmap_set_max_osd(map, max); 1436 if (err) 1437 goto bad; 1438 } 1439 1440 map->epoch++; 1441 map->modified = modified; 1442 if (newcrush) { 1443 if (map->crush) 1444 crush_destroy(map->crush); 1445 map->crush = newcrush; 1446 newcrush = NULL; 1447 } 1448 1449 /* new_pools */ 1450 err = decode_new_pools(p, end, map); 1451 if (err) 1452 goto bad; 1453 1454 /* new_pool_names */ 1455 err = decode_pool_names(p, end, map); 1456 if (err) 1457 goto bad; 1458 1459 /* old_pool */ 1460 ceph_decode_32_safe(p, end, len, e_inval); 1461 while (len--) { 1462 struct ceph_pg_pool_info *pi; 1463 1464 ceph_decode_64_safe(p, end, pool, e_inval); 1465 pi = __lookup_pg_pool(&map->pg_pools, pool); 1466 if (pi) 1467 __remove_pg_pool(&map->pg_pools, pi); 1468 } 1469 1470 /* new_up_client, new_state, new_weight */ 1471 err = decode_new_up_state_weight(p, end, map); 1472 if (err) 1473 goto bad; 1474 1475 /* new_pg_temp */ 1476 err = decode_new_pg_temp(p, end, map); 1477 if (err) 1478 goto bad; 1479 1480 /* new_primary_temp */ 1481 if (struct_v >= 1) { 1482 err = decode_new_primary_temp(p, end, map); 1483 if (err) 1484 goto bad; 1485 } 1486 1487 /* new_primary_affinity */ 1488 if (struct_v >= 2) { 1489 err = decode_new_primary_affinity(p, end, map); 1490 if (err) 1491 goto bad; 1492 } 1493 1494 /* ignore the rest */ 1495 *p = end; 1496 1497 dout("inc osdmap epoch %d max_osd %d\n", map->epoch, map->max_osd); 1498 return map; 1499 1500 e_inval: 1501 err = -EINVAL; 1502 bad: 1503 pr_err("corrupt inc osdmap (%d) epoch %d off %d (%p of %p-%p)\n", 1504 err, epoch, (int)(*p - start), *p, start, end); 1505 print_hex_dump(KERN_DEBUG, "osdmap: ", 1506 DUMP_PREFIX_OFFSET, 16, 1, 1507 start, end - start, true); 1508 if (newcrush) 1509 crush_destroy(newcrush); 1510 return ERR_PTR(err); 1511 } 1512 1513 void ceph_oid_copy(struct ceph_object_id *dest, 1514 const struct ceph_object_id *src) 1515 { 1516 WARN_ON(!ceph_oid_empty(dest)); 1517 1518 if (src->name != src->inline_name) { 1519 /* very rare, see ceph_object_id definition */ 1520 dest->name = kmalloc(src->name_len + 1, 1521 GFP_NOIO | __GFP_NOFAIL); 1522 } 1523 1524 memcpy(dest->name, src->name, src->name_len + 1); 1525 dest->name_len = src->name_len; 1526 } 1527 EXPORT_SYMBOL(ceph_oid_copy); 1528 1529 static __printf(2, 0) 1530 int oid_printf_vargs(struct ceph_object_id *oid, const char *fmt, va_list ap) 1531 { 1532 int len; 1533 1534 WARN_ON(!ceph_oid_empty(oid)); 1535 1536 len = vsnprintf(oid->inline_name, sizeof(oid->inline_name), fmt, ap); 1537 if (len >= sizeof(oid->inline_name)) 1538 return len; 1539 1540 oid->name_len = len; 1541 return 0; 1542 } 1543 1544 /* 1545 * If oid doesn't fit into inline buffer, BUG. 1546 */ 1547 void ceph_oid_printf(struct ceph_object_id *oid, const char *fmt, ...) 1548 { 1549 va_list ap; 1550 1551 va_start(ap, fmt); 1552 BUG_ON(oid_printf_vargs(oid, fmt, ap)); 1553 va_end(ap); 1554 } 1555 EXPORT_SYMBOL(ceph_oid_printf); 1556 1557 static __printf(3, 0) 1558 int oid_aprintf_vargs(struct ceph_object_id *oid, gfp_t gfp, 1559 const char *fmt, va_list ap) 1560 { 1561 va_list aq; 1562 int len; 1563 1564 va_copy(aq, ap); 1565 len = oid_printf_vargs(oid, fmt, aq); 1566 va_end(aq); 1567 1568 if (len) { 1569 char *external_name; 1570 1571 external_name = kmalloc(len + 1, gfp); 1572 if (!external_name) 1573 return -ENOMEM; 1574 1575 oid->name = external_name; 1576 WARN_ON(vsnprintf(oid->name, len + 1, fmt, ap) != len); 1577 oid->name_len = len; 1578 } 1579 1580 return 0; 1581 } 1582 1583 /* 1584 * If oid doesn't fit into inline buffer, allocate. 1585 */ 1586 int ceph_oid_aprintf(struct ceph_object_id *oid, gfp_t gfp, 1587 const char *fmt, ...) 1588 { 1589 va_list ap; 1590 int ret; 1591 1592 va_start(ap, fmt); 1593 ret = oid_aprintf_vargs(oid, gfp, fmt, ap); 1594 va_end(ap); 1595 1596 return ret; 1597 } 1598 EXPORT_SYMBOL(ceph_oid_aprintf); 1599 1600 void ceph_oid_destroy(struct ceph_object_id *oid) 1601 { 1602 if (oid->name != oid->inline_name) 1603 kfree(oid->name); 1604 } 1605 EXPORT_SYMBOL(ceph_oid_destroy); 1606 1607 /* 1608 * osds only 1609 */ 1610 static bool __osds_equal(const struct ceph_osds *lhs, 1611 const struct ceph_osds *rhs) 1612 { 1613 if (lhs->size == rhs->size && 1614 !memcmp(lhs->osds, rhs->osds, rhs->size * sizeof(rhs->osds[0]))) 1615 return true; 1616 1617 return false; 1618 } 1619 1620 /* 1621 * osds + primary 1622 */ 1623 static bool osds_equal(const struct ceph_osds *lhs, 1624 const struct ceph_osds *rhs) 1625 { 1626 if (__osds_equal(lhs, rhs) && 1627 lhs->primary == rhs->primary) 1628 return true; 1629 1630 return false; 1631 } 1632 1633 static bool osds_valid(const struct ceph_osds *set) 1634 { 1635 /* non-empty set */ 1636 if (set->size > 0 && set->primary >= 0) 1637 return true; 1638 1639 /* empty can_shift_osds set */ 1640 if (!set->size && set->primary == -1) 1641 return true; 1642 1643 /* empty !can_shift_osds set - all NONE */ 1644 if (set->size > 0 && set->primary == -1) { 1645 int i; 1646 1647 for (i = 0; i < set->size; i++) { 1648 if (set->osds[i] != CRUSH_ITEM_NONE) 1649 break; 1650 } 1651 if (i == set->size) 1652 return true; 1653 } 1654 1655 return false; 1656 } 1657 1658 void ceph_osds_copy(struct ceph_osds *dest, const struct ceph_osds *src) 1659 { 1660 memcpy(dest->osds, src->osds, src->size * sizeof(src->osds[0])); 1661 dest->size = src->size; 1662 dest->primary = src->primary; 1663 } 1664 1665 static bool is_split(const struct ceph_pg *pgid, 1666 u32 old_pg_num, 1667 u32 new_pg_num) 1668 { 1669 int old_bits = calc_bits_of(old_pg_num); 1670 int old_mask = (1 << old_bits) - 1; 1671 int n; 1672 1673 WARN_ON(pgid->seed >= old_pg_num); 1674 if (new_pg_num <= old_pg_num) 1675 return false; 1676 1677 for (n = 1; ; n++) { 1678 int next_bit = n << (old_bits - 1); 1679 u32 s = next_bit | pgid->seed; 1680 1681 if (s < old_pg_num || s == pgid->seed) 1682 continue; 1683 if (s >= new_pg_num) 1684 break; 1685 1686 s = ceph_stable_mod(s, old_pg_num, old_mask); 1687 if (s == pgid->seed) 1688 return true; 1689 } 1690 1691 return false; 1692 } 1693 1694 bool ceph_is_new_interval(const struct ceph_osds *old_acting, 1695 const struct ceph_osds *new_acting, 1696 const struct ceph_osds *old_up, 1697 const struct ceph_osds *new_up, 1698 int old_size, 1699 int new_size, 1700 int old_min_size, 1701 int new_min_size, 1702 u32 old_pg_num, 1703 u32 new_pg_num, 1704 bool old_sort_bitwise, 1705 bool new_sort_bitwise, 1706 const struct ceph_pg *pgid) 1707 { 1708 return !osds_equal(old_acting, new_acting) || 1709 !osds_equal(old_up, new_up) || 1710 old_size != new_size || 1711 old_min_size != new_min_size || 1712 is_split(pgid, old_pg_num, new_pg_num) || 1713 old_sort_bitwise != new_sort_bitwise; 1714 } 1715 1716 static int calc_pg_rank(int osd, const struct ceph_osds *acting) 1717 { 1718 int i; 1719 1720 for (i = 0; i < acting->size; i++) { 1721 if (acting->osds[i] == osd) 1722 return i; 1723 } 1724 1725 return -1; 1726 } 1727 1728 static bool primary_changed(const struct ceph_osds *old_acting, 1729 const struct ceph_osds *new_acting) 1730 { 1731 if (!old_acting->size && !new_acting->size) 1732 return false; /* both still empty */ 1733 1734 if (!old_acting->size ^ !new_acting->size) 1735 return true; /* was empty, now not, or vice versa */ 1736 1737 if (old_acting->primary != new_acting->primary) 1738 return true; /* primary changed */ 1739 1740 if (calc_pg_rank(old_acting->primary, old_acting) != 1741 calc_pg_rank(new_acting->primary, new_acting)) 1742 return true; 1743 1744 return false; /* same primary (tho replicas may have changed) */ 1745 } 1746 1747 bool ceph_osds_changed(const struct ceph_osds *old_acting, 1748 const struct ceph_osds *new_acting, 1749 bool any_change) 1750 { 1751 if (primary_changed(old_acting, new_acting)) 1752 return true; 1753 1754 if (any_change && !__osds_equal(old_acting, new_acting)) 1755 return true; 1756 1757 return false; 1758 } 1759 1760 /* 1761 * calculate file layout from given offset, length. 1762 * fill in correct oid, logical length, and object extent 1763 * offset, length. 1764 * 1765 * for now, we write only a single su, until we can 1766 * pass a stride back to the caller. 1767 */ 1768 int ceph_calc_file_object_mapping(struct ceph_file_layout *layout, 1769 u64 off, u64 len, 1770 u64 *ono, 1771 u64 *oxoff, u64 *oxlen) 1772 { 1773 u32 osize = layout->object_size; 1774 u32 su = layout->stripe_unit; 1775 u32 sc = layout->stripe_count; 1776 u32 bl, stripeno, stripepos, objsetno; 1777 u32 su_per_object; 1778 u64 t, su_offset; 1779 1780 dout("mapping %llu~%llu osize %u fl_su %u\n", off, len, 1781 osize, su); 1782 if (su == 0 || sc == 0) 1783 goto invalid; 1784 su_per_object = osize / su; 1785 if (su_per_object == 0) 1786 goto invalid; 1787 dout("osize %u / su %u = su_per_object %u\n", osize, su, 1788 su_per_object); 1789 1790 if ((su & ~PAGE_MASK) != 0) 1791 goto invalid; 1792 1793 /* bl = *off / su; */ 1794 t = off; 1795 do_div(t, su); 1796 bl = t; 1797 dout("off %llu / su %u = bl %u\n", off, su, bl); 1798 1799 stripeno = bl / sc; 1800 stripepos = bl % sc; 1801 objsetno = stripeno / su_per_object; 1802 1803 *ono = objsetno * sc + stripepos; 1804 dout("objset %u * sc %u = ono %u\n", objsetno, sc, (unsigned int)*ono); 1805 1806 /* *oxoff = *off % layout->fl_stripe_unit; # offset in su */ 1807 t = off; 1808 su_offset = do_div(t, su); 1809 *oxoff = su_offset + (stripeno % su_per_object) * su; 1810 1811 /* 1812 * Calculate the length of the extent being written to the selected 1813 * object. This is the minimum of the full length requested (len) or 1814 * the remainder of the current stripe being written to. 1815 */ 1816 *oxlen = min_t(u64, len, su - su_offset); 1817 1818 dout(" obj extent %llu~%llu\n", *oxoff, *oxlen); 1819 return 0; 1820 1821 invalid: 1822 dout(" invalid layout\n"); 1823 *ono = 0; 1824 *oxoff = 0; 1825 *oxlen = 0; 1826 return -EINVAL; 1827 } 1828 EXPORT_SYMBOL(ceph_calc_file_object_mapping); 1829 1830 /* 1831 * Map an object into a PG. 1832 * 1833 * Should only be called with target_oid and target_oloc (as opposed to 1834 * base_oid and base_oloc), since tiering isn't taken into account. 1835 */ 1836 int ceph_object_locator_to_pg(struct ceph_osdmap *osdmap, 1837 struct ceph_object_id *oid, 1838 struct ceph_object_locator *oloc, 1839 struct ceph_pg *raw_pgid) 1840 { 1841 struct ceph_pg_pool_info *pi; 1842 1843 pi = ceph_pg_pool_by_id(osdmap, oloc->pool); 1844 if (!pi) 1845 return -ENOENT; 1846 1847 raw_pgid->pool = oloc->pool; 1848 raw_pgid->seed = ceph_str_hash(pi->object_hash, oid->name, 1849 oid->name_len); 1850 1851 dout("%s %s -> raw_pgid %llu.%x\n", __func__, oid->name, 1852 raw_pgid->pool, raw_pgid->seed); 1853 return 0; 1854 } 1855 EXPORT_SYMBOL(ceph_object_locator_to_pg); 1856 1857 /* 1858 * Map a raw PG (full precision ps) into an actual PG. 1859 */ 1860 static void raw_pg_to_pg(struct ceph_pg_pool_info *pi, 1861 const struct ceph_pg *raw_pgid, 1862 struct ceph_pg *pgid) 1863 { 1864 pgid->pool = raw_pgid->pool; 1865 pgid->seed = ceph_stable_mod(raw_pgid->seed, pi->pg_num, 1866 pi->pg_num_mask); 1867 } 1868 1869 /* 1870 * Map a raw PG (full precision ps) into a placement ps (placement 1871 * seed). Include pool id in that value so that different pools don't 1872 * use the same seeds. 1873 */ 1874 static u32 raw_pg_to_pps(struct ceph_pg_pool_info *pi, 1875 const struct ceph_pg *raw_pgid) 1876 { 1877 if (pi->flags & CEPH_POOL_FLAG_HASHPSPOOL) { 1878 /* hash pool id and seed so that pool PGs do not overlap */ 1879 return crush_hash32_2(CRUSH_HASH_RJENKINS1, 1880 ceph_stable_mod(raw_pgid->seed, 1881 pi->pgp_num, 1882 pi->pgp_num_mask), 1883 raw_pgid->pool); 1884 } else { 1885 /* 1886 * legacy behavior: add ps and pool together. this is 1887 * not a great approach because the PGs from each pool 1888 * will overlap on top of each other: 0.5 == 1.4 == 1889 * 2.3 == ... 1890 */ 1891 return ceph_stable_mod(raw_pgid->seed, pi->pgp_num, 1892 pi->pgp_num_mask) + 1893 (unsigned)raw_pgid->pool; 1894 } 1895 } 1896 1897 static int do_crush(struct ceph_osdmap *map, int ruleno, int x, 1898 int *result, int result_max, 1899 const __u32 *weight, int weight_max) 1900 { 1901 int r; 1902 1903 BUG_ON(result_max > CEPH_PG_MAX_SIZE); 1904 1905 mutex_lock(&map->crush_scratch_mutex); 1906 r = crush_do_rule(map->crush, ruleno, x, result, result_max, 1907 weight, weight_max, map->crush_scratch_ary); 1908 mutex_unlock(&map->crush_scratch_mutex); 1909 1910 return r; 1911 } 1912 1913 /* 1914 * Calculate raw set (CRUSH output) for given PG. The result may 1915 * contain nonexistent OSDs. ->primary is undefined for a raw set. 1916 * 1917 * Placement seed (CRUSH input) is returned through @ppps. 1918 */ 1919 static void pg_to_raw_osds(struct ceph_osdmap *osdmap, 1920 struct ceph_pg_pool_info *pi, 1921 const struct ceph_pg *raw_pgid, 1922 struct ceph_osds *raw, 1923 u32 *ppps) 1924 { 1925 u32 pps = raw_pg_to_pps(pi, raw_pgid); 1926 int ruleno; 1927 int len; 1928 1929 ceph_osds_init(raw); 1930 if (ppps) 1931 *ppps = pps; 1932 1933 ruleno = crush_find_rule(osdmap->crush, pi->crush_ruleset, pi->type, 1934 pi->size); 1935 if (ruleno < 0) { 1936 pr_err("no crush rule: pool %lld ruleset %d type %d size %d\n", 1937 pi->id, pi->crush_ruleset, pi->type, pi->size); 1938 return; 1939 } 1940 1941 len = do_crush(osdmap, ruleno, pps, raw->osds, 1942 min_t(int, pi->size, ARRAY_SIZE(raw->osds)), 1943 osdmap->osd_weight, osdmap->max_osd); 1944 if (len < 0) { 1945 pr_err("error %d from crush rule %d: pool %lld ruleset %d type %d size %d\n", 1946 len, ruleno, pi->id, pi->crush_ruleset, pi->type, 1947 pi->size); 1948 return; 1949 } 1950 1951 raw->size = len; 1952 } 1953 1954 /* 1955 * Given raw set, calculate up set and up primary. By definition of an 1956 * up set, the result won't contain nonexistent or down OSDs. 1957 * 1958 * This is done in-place - on return @set is the up set. If it's 1959 * empty, ->primary will remain undefined. 1960 */ 1961 static void raw_to_up_osds(struct ceph_osdmap *osdmap, 1962 struct ceph_pg_pool_info *pi, 1963 struct ceph_osds *set) 1964 { 1965 int i; 1966 1967 /* ->primary is undefined for a raw set */ 1968 BUG_ON(set->primary != -1); 1969 1970 if (ceph_can_shift_osds(pi)) { 1971 int removed = 0; 1972 1973 /* shift left */ 1974 for (i = 0; i < set->size; i++) { 1975 if (ceph_osd_is_down(osdmap, set->osds[i])) { 1976 removed++; 1977 continue; 1978 } 1979 if (removed) 1980 set->osds[i - removed] = set->osds[i]; 1981 } 1982 set->size -= removed; 1983 if (set->size > 0) 1984 set->primary = set->osds[0]; 1985 } else { 1986 /* set down/dne devices to NONE */ 1987 for (i = set->size - 1; i >= 0; i--) { 1988 if (ceph_osd_is_down(osdmap, set->osds[i])) 1989 set->osds[i] = CRUSH_ITEM_NONE; 1990 else 1991 set->primary = set->osds[i]; 1992 } 1993 } 1994 } 1995 1996 static void apply_primary_affinity(struct ceph_osdmap *osdmap, 1997 struct ceph_pg_pool_info *pi, 1998 u32 pps, 1999 struct ceph_osds *up) 2000 { 2001 int i; 2002 int pos = -1; 2003 2004 /* 2005 * Do we have any non-default primary_affinity values for these 2006 * osds? 2007 */ 2008 if (!osdmap->osd_primary_affinity) 2009 return; 2010 2011 for (i = 0; i < up->size; i++) { 2012 int osd = up->osds[i]; 2013 2014 if (osd != CRUSH_ITEM_NONE && 2015 osdmap->osd_primary_affinity[osd] != 2016 CEPH_OSD_DEFAULT_PRIMARY_AFFINITY) { 2017 break; 2018 } 2019 } 2020 if (i == up->size) 2021 return; 2022 2023 /* 2024 * Pick the primary. Feed both the seed (for the pg) and the 2025 * osd into the hash/rng so that a proportional fraction of an 2026 * osd's pgs get rejected as primary. 2027 */ 2028 for (i = 0; i < up->size; i++) { 2029 int osd = up->osds[i]; 2030 u32 aff; 2031 2032 if (osd == CRUSH_ITEM_NONE) 2033 continue; 2034 2035 aff = osdmap->osd_primary_affinity[osd]; 2036 if (aff < CEPH_OSD_MAX_PRIMARY_AFFINITY && 2037 (crush_hash32_2(CRUSH_HASH_RJENKINS1, 2038 pps, osd) >> 16) >= aff) { 2039 /* 2040 * We chose not to use this primary. Note it 2041 * anyway as a fallback in case we don't pick 2042 * anyone else, but keep looking. 2043 */ 2044 if (pos < 0) 2045 pos = i; 2046 } else { 2047 pos = i; 2048 break; 2049 } 2050 } 2051 if (pos < 0) 2052 return; 2053 2054 up->primary = up->osds[pos]; 2055 2056 if (ceph_can_shift_osds(pi) && pos > 0) { 2057 /* move the new primary to the front */ 2058 for (i = pos; i > 0; i--) 2059 up->osds[i] = up->osds[i - 1]; 2060 up->osds[0] = up->primary; 2061 } 2062 } 2063 2064 /* 2065 * Get pg_temp and primary_temp mappings for given PG. 2066 * 2067 * Note that a PG may have none, only pg_temp, only primary_temp or 2068 * both pg_temp and primary_temp mappings. This means @temp isn't 2069 * always a valid OSD set on return: in the "only primary_temp" case, 2070 * @temp will have its ->primary >= 0 but ->size == 0. 2071 */ 2072 static void get_temp_osds(struct ceph_osdmap *osdmap, 2073 struct ceph_pg_pool_info *pi, 2074 const struct ceph_pg *raw_pgid, 2075 struct ceph_osds *temp) 2076 { 2077 struct ceph_pg pgid; 2078 struct ceph_pg_mapping *pg; 2079 int i; 2080 2081 raw_pg_to_pg(pi, raw_pgid, &pgid); 2082 ceph_osds_init(temp); 2083 2084 /* pg_temp? */ 2085 pg = __lookup_pg_mapping(&osdmap->pg_temp, pgid); 2086 if (pg) { 2087 for (i = 0; i < pg->pg_temp.len; i++) { 2088 if (ceph_osd_is_down(osdmap, pg->pg_temp.osds[i])) { 2089 if (ceph_can_shift_osds(pi)) 2090 continue; 2091 2092 temp->osds[temp->size++] = CRUSH_ITEM_NONE; 2093 } else { 2094 temp->osds[temp->size++] = pg->pg_temp.osds[i]; 2095 } 2096 } 2097 2098 /* apply pg_temp's primary */ 2099 for (i = 0; i < temp->size; i++) { 2100 if (temp->osds[i] != CRUSH_ITEM_NONE) { 2101 temp->primary = temp->osds[i]; 2102 break; 2103 } 2104 } 2105 } 2106 2107 /* primary_temp? */ 2108 pg = __lookup_pg_mapping(&osdmap->primary_temp, pgid); 2109 if (pg) 2110 temp->primary = pg->primary_temp.osd; 2111 } 2112 2113 /* 2114 * Map a PG to its acting set as well as its up set. 2115 * 2116 * Acting set is used for data mapping purposes, while up set can be 2117 * recorded for detecting interval changes and deciding whether to 2118 * resend a request. 2119 */ 2120 void ceph_pg_to_up_acting_osds(struct ceph_osdmap *osdmap, 2121 const struct ceph_pg *raw_pgid, 2122 struct ceph_osds *up, 2123 struct ceph_osds *acting) 2124 { 2125 struct ceph_pg_pool_info *pi; 2126 u32 pps; 2127 2128 pi = ceph_pg_pool_by_id(osdmap, raw_pgid->pool); 2129 if (!pi) { 2130 ceph_osds_init(up); 2131 ceph_osds_init(acting); 2132 goto out; 2133 } 2134 2135 pg_to_raw_osds(osdmap, pi, raw_pgid, up, &pps); 2136 raw_to_up_osds(osdmap, pi, up); 2137 apply_primary_affinity(osdmap, pi, pps, up); 2138 get_temp_osds(osdmap, pi, raw_pgid, acting); 2139 if (!acting->size) { 2140 memcpy(acting->osds, up->osds, up->size * sizeof(up->osds[0])); 2141 acting->size = up->size; 2142 if (acting->primary == -1) 2143 acting->primary = up->primary; 2144 } 2145 out: 2146 WARN_ON(!osds_valid(up) || !osds_valid(acting)); 2147 } 2148 2149 /* 2150 * Return acting primary for given PG, or -1 if none. 2151 */ 2152 int ceph_pg_to_acting_primary(struct ceph_osdmap *osdmap, 2153 const struct ceph_pg *raw_pgid) 2154 { 2155 struct ceph_osds up, acting; 2156 2157 ceph_pg_to_up_acting_osds(osdmap, raw_pgid, &up, &acting); 2158 return acting.primary; 2159 } 2160 EXPORT_SYMBOL(ceph_pg_to_acting_primary); 2161