1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2010 Kent Overstreet <kent.overstreet@gmail.com> 4 * Copyright (C) 2014 Datera Inc. 5 */ 6 7 #include "bcachefs.h" 8 #include "alloc_background.h" 9 #include "alloc_foreground.h" 10 #include "backpointers.h" 11 #include "bkey_methods.h" 12 #include "bkey_buf.h" 13 #include "btree_journal_iter.h" 14 #include "btree_key_cache.h" 15 #include "btree_locking.h" 16 #include "btree_node_scan.h" 17 #include "btree_update_interior.h" 18 #include "btree_io.h" 19 #include "btree_gc.h" 20 #include "buckets.h" 21 #include "clock.h" 22 #include "debug.h" 23 #include "ec.h" 24 #include "error.h" 25 #include "extents.h" 26 #include "journal.h" 27 #include "keylist.h" 28 #include "move.h" 29 #include "recovery_passes.h" 30 #include "reflink.h" 31 #include "replicas.h" 32 #include "super-io.h" 33 #include "trace.h" 34 35 #include <linux/slab.h> 36 #include <linux/bitops.h> 37 #include <linux/freezer.h> 38 #include <linux/kthread.h> 39 #include <linux/preempt.h> 40 #include <linux/rcupdate.h> 41 #include <linux/sched/task.h> 42 43 #define DROP_THIS_NODE 10 44 #define DROP_PREV_NODE 11 45 #define DID_FILL_FROM_SCAN 12 46 47 static struct bkey_s unsafe_bkey_s_c_to_s(struct bkey_s_c k) 48 { 49 return (struct bkey_s) {{{ 50 (struct bkey *) k.k, 51 (struct bch_val *) k.v 52 }}}; 53 } 54 55 static bool should_restart_for_topology_repair(struct bch_fs *c) 56 { 57 return c->opts.fix_errors != FSCK_FIX_no && 58 !(c->recovery_passes_complete & BIT_ULL(BCH_RECOVERY_PASS_check_topology)); 59 } 60 61 static inline void __gc_pos_set(struct bch_fs *c, struct gc_pos new_pos) 62 { 63 preempt_disable(); 64 write_seqcount_begin(&c->gc_pos_lock); 65 c->gc_pos = new_pos; 66 write_seqcount_end(&c->gc_pos_lock); 67 preempt_enable(); 68 } 69 70 static inline void gc_pos_set(struct bch_fs *c, struct gc_pos new_pos) 71 { 72 BUG_ON(gc_pos_cmp(new_pos, c->gc_pos) <= 0); 73 __gc_pos_set(c, new_pos); 74 } 75 76 static void btree_ptr_to_v2(struct btree *b, struct bkey_i_btree_ptr_v2 *dst) 77 { 78 switch (b->key.k.type) { 79 case KEY_TYPE_btree_ptr: { 80 struct bkey_i_btree_ptr *src = bkey_i_to_btree_ptr(&b->key); 81 82 dst->k.p = src->k.p; 83 dst->v.mem_ptr = 0; 84 dst->v.seq = b->data->keys.seq; 85 dst->v.sectors_written = 0; 86 dst->v.flags = 0; 87 dst->v.min_key = b->data->min_key; 88 set_bkey_val_bytes(&dst->k, sizeof(dst->v) + bkey_val_bytes(&src->k)); 89 memcpy(dst->v.start, src->v.start, bkey_val_bytes(&src->k)); 90 break; 91 } 92 case KEY_TYPE_btree_ptr_v2: 93 bkey_copy(&dst->k_i, &b->key); 94 break; 95 default: 96 BUG(); 97 } 98 } 99 100 static void bch2_btree_node_update_key_early(struct btree_trans *trans, 101 enum btree_id btree, unsigned level, 102 struct bkey_s_c old, struct bkey_i *new) 103 { 104 struct bch_fs *c = trans->c; 105 struct btree *b; 106 struct bkey_buf tmp; 107 int ret; 108 109 bch2_bkey_buf_init(&tmp); 110 bch2_bkey_buf_reassemble(&tmp, c, old); 111 112 b = bch2_btree_node_get_noiter(trans, tmp.k, btree, level, true); 113 if (!IS_ERR_OR_NULL(b)) { 114 mutex_lock(&c->btree_cache.lock); 115 116 bch2_btree_node_hash_remove(&c->btree_cache, b); 117 118 bkey_copy(&b->key, new); 119 ret = __bch2_btree_node_hash_insert(&c->btree_cache, b); 120 BUG_ON(ret); 121 122 mutex_unlock(&c->btree_cache.lock); 123 six_unlock_read(&b->c.lock); 124 } 125 126 bch2_bkey_buf_exit(&tmp, c); 127 } 128 129 static int set_node_min(struct bch_fs *c, struct btree *b, struct bpos new_min) 130 { 131 struct bkey_i_btree_ptr_v2 *new; 132 int ret; 133 134 if (c->opts.verbose) { 135 struct printbuf buf = PRINTBUF; 136 137 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); 138 prt_str(&buf, " -> "); 139 bch2_bpos_to_text(&buf, new_min); 140 141 bch_info(c, "%s(): %s", __func__, buf.buf); 142 printbuf_exit(&buf); 143 } 144 145 new = kmalloc_array(BKEY_BTREE_PTR_U64s_MAX, sizeof(u64), GFP_KERNEL); 146 if (!new) 147 return -BCH_ERR_ENOMEM_gc_repair_key; 148 149 btree_ptr_to_v2(b, new); 150 b->data->min_key = new_min; 151 new->v.min_key = new_min; 152 SET_BTREE_PTR_RANGE_UPDATED(&new->v, true); 153 154 ret = bch2_journal_key_insert_take(c, b->c.btree_id, b->c.level + 1, &new->k_i); 155 if (ret) { 156 kfree(new); 157 return ret; 158 } 159 160 bch2_btree_node_drop_keys_outside_node(b); 161 bkey_copy(&b->key, &new->k_i); 162 return 0; 163 } 164 165 static int set_node_max(struct bch_fs *c, struct btree *b, struct bpos new_max) 166 { 167 struct bkey_i_btree_ptr_v2 *new; 168 int ret; 169 170 if (c->opts.verbose) { 171 struct printbuf buf = PRINTBUF; 172 173 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); 174 prt_str(&buf, " -> "); 175 bch2_bpos_to_text(&buf, new_max); 176 177 bch_info(c, "%s(): %s", __func__, buf.buf); 178 printbuf_exit(&buf); 179 } 180 181 ret = bch2_journal_key_delete(c, b->c.btree_id, b->c.level + 1, b->key.k.p); 182 if (ret) 183 return ret; 184 185 new = kmalloc_array(BKEY_BTREE_PTR_U64s_MAX, sizeof(u64), GFP_KERNEL); 186 if (!new) 187 return -BCH_ERR_ENOMEM_gc_repair_key; 188 189 btree_ptr_to_v2(b, new); 190 b->data->max_key = new_max; 191 new->k.p = new_max; 192 SET_BTREE_PTR_RANGE_UPDATED(&new->v, true); 193 194 ret = bch2_journal_key_insert_take(c, b->c.btree_id, b->c.level + 1, &new->k_i); 195 if (ret) { 196 kfree(new); 197 return ret; 198 } 199 200 bch2_btree_node_drop_keys_outside_node(b); 201 202 mutex_lock(&c->btree_cache.lock); 203 bch2_btree_node_hash_remove(&c->btree_cache, b); 204 205 bkey_copy(&b->key, &new->k_i); 206 ret = __bch2_btree_node_hash_insert(&c->btree_cache, b); 207 BUG_ON(ret); 208 mutex_unlock(&c->btree_cache.lock); 209 return 0; 210 } 211 212 static int btree_check_node_boundaries(struct bch_fs *c, struct btree *b, 213 struct btree *prev, struct btree *cur, 214 struct bpos *pulled_from_scan) 215 { 216 struct bpos expected_start = !prev 217 ? b->data->min_key 218 : bpos_successor(prev->key.k.p); 219 struct printbuf buf = PRINTBUF; 220 int ret = 0; 221 222 BUG_ON(b->key.k.type == KEY_TYPE_btree_ptr_v2 && 223 !bpos_eq(bkey_i_to_btree_ptr_v2(&b->key)->v.min_key, 224 b->data->min_key)); 225 226 if (bpos_eq(expected_start, cur->data->min_key)) 227 return 0; 228 229 prt_printf(&buf, " at btree %s level %u:\n parent: ", 230 bch2_btree_id_str(b->c.btree_id), b->c.level); 231 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); 232 233 if (prev) { 234 prt_printf(&buf, "\n prev: "); 235 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&prev->key)); 236 } 237 238 prt_str(&buf, "\n next: "); 239 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&cur->key)); 240 241 if (bpos_lt(expected_start, cur->data->min_key)) { /* gap */ 242 if (b->c.level == 1 && 243 bpos_lt(*pulled_from_scan, cur->data->min_key)) { 244 ret = bch2_get_scanned_nodes(c, b->c.btree_id, 0, 245 expected_start, 246 bpos_predecessor(cur->data->min_key)); 247 if (ret) 248 goto err; 249 250 *pulled_from_scan = cur->data->min_key; 251 ret = DID_FILL_FROM_SCAN; 252 } else { 253 if (mustfix_fsck_err(c, btree_node_topology_bad_min_key, 254 "btree node with incorrect min_key%s", buf.buf)) 255 ret = set_node_min(c, cur, expected_start); 256 } 257 } else { /* overlap */ 258 if (prev && BTREE_NODE_SEQ(cur->data) > BTREE_NODE_SEQ(prev->data)) { /* cur overwrites prev */ 259 if (bpos_ge(prev->data->min_key, cur->data->min_key)) { /* fully? */ 260 if (mustfix_fsck_err(c, btree_node_topology_overwritten_by_next_node, 261 "btree node overwritten by next node%s", buf.buf)) 262 ret = DROP_PREV_NODE; 263 } else { 264 if (mustfix_fsck_err(c, btree_node_topology_bad_max_key, 265 "btree node with incorrect max_key%s", buf.buf)) 266 ret = set_node_max(c, prev, 267 bpos_predecessor(cur->data->min_key)); 268 } 269 } else { 270 if (bpos_ge(expected_start, cur->data->max_key)) { /* fully? */ 271 if (mustfix_fsck_err(c, btree_node_topology_overwritten_by_prev_node, 272 "btree node overwritten by prev node%s", buf.buf)) 273 ret = DROP_THIS_NODE; 274 } else { 275 if (mustfix_fsck_err(c, btree_node_topology_bad_min_key, 276 "btree node with incorrect min_key%s", buf.buf)) 277 ret = set_node_min(c, cur, expected_start); 278 } 279 } 280 } 281 err: 282 fsck_err: 283 printbuf_exit(&buf); 284 return ret; 285 } 286 287 static int btree_repair_node_end(struct bch_fs *c, struct btree *b, 288 struct btree *child, struct bpos *pulled_from_scan) 289 { 290 struct printbuf buf = PRINTBUF; 291 int ret = 0; 292 293 if (bpos_eq(child->key.k.p, b->key.k.p)) 294 return 0; 295 296 prt_printf(&buf, "at btree %s level %u:\n parent: ", 297 bch2_btree_id_str(b->c.btree_id), b->c.level); 298 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); 299 300 prt_str(&buf, "\n child: "); 301 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&child->key)); 302 303 if (mustfix_fsck_err(c, btree_node_topology_bad_max_key, 304 "btree node with incorrect max_key%s", buf.buf)) { 305 if (b->c.level == 1 && 306 bpos_lt(*pulled_from_scan, b->key.k.p)) { 307 ret = bch2_get_scanned_nodes(c, b->c.btree_id, 0, 308 bpos_successor(child->key.k.p), b->key.k.p); 309 if (ret) 310 goto err; 311 312 *pulled_from_scan = b->key.k.p; 313 ret = DID_FILL_FROM_SCAN; 314 } else { 315 ret = set_node_max(c, child, b->key.k.p); 316 } 317 } 318 err: 319 fsck_err: 320 printbuf_exit(&buf); 321 return ret; 322 } 323 324 static int bch2_btree_repair_topology_recurse(struct btree_trans *trans, struct btree *b, 325 struct bpos *pulled_from_scan) 326 { 327 struct bch_fs *c = trans->c; 328 struct btree_and_journal_iter iter; 329 struct bkey_s_c k; 330 struct bkey_buf prev_k, cur_k; 331 struct btree *prev = NULL, *cur = NULL; 332 bool have_child, new_pass = false; 333 struct printbuf buf = PRINTBUF; 334 int ret = 0; 335 336 if (!b->c.level) 337 return 0; 338 339 bch2_bkey_buf_init(&prev_k); 340 bch2_bkey_buf_init(&cur_k); 341 again: 342 cur = prev = NULL; 343 have_child = new_pass = false; 344 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 345 iter.prefetch = true; 346 347 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 348 BUG_ON(bpos_lt(k.k->p, b->data->min_key)); 349 BUG_ON(bpos_gt(k.k->p, b->data->max_key)); 350 351 bch2_btree_and_journal_iter_advance(&iter); 352 bch2_bkey_buf_reassemble(&cur_k, c, k); 353 354 cur = bch2_btree_node_get_noiter(trans, cur_k.k, 355 b->c.btree_id, b->c.level - 1, 356 false); 357 ret = PTR_ERR_OR_ZERO(cur); 358 359 printbuf_reset(&buf); 360 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur_k.k)); 361 362 if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO), c, 363 btree_node_unreadable, 364 "Topology repair: unreadable btree node at btree %s level %u:\n" 365 " %s", 366 bch2_btree_id_str(b->c.btree_id), 367 b->c.level - 1, 368 buf.buf)) { 369 bch2_btree_node_evict(trans, cur_k.k); 370 cur = NULL; 371 ret = bch2_journal_key_delete(c, b->c.btree_id, 372 b->c.level, cur_k.k->k.p); 373 if (ret) 374 break; 375 376 if (!btree_id_is_alloc(b->c.btree_id)) { 377 ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes); 378 if (ret) 379 break; 380 } 381 continue; 382 } 383 384 bch_err_msg(c, ret, "getting btree node"); 385 if (ret) 386 break; 387 388 if (bch2_btree_node_is_stale(c, cur)) { 389 bch_info(c, "btree node %s older than nodes found by scanning", buf.buf); 390 six_unlock_read(&cur->c.lock); 391 bch2_btree_node_evict(trans, cur_k.k); 392 ret = bch2_journal_key_delete(c, b->c.btree_id, 393 b->c.level, cur_k.k->k.p); 394 cur = NULL; 395 if (ret) 396 break; 397 continue; 398 } 399 400 ret = btree_check_node_boundaries(c, b, prev, cur, pulled_from_scan); 401 if (ret == DID_FILL_FROM_SCAN) { 402 new_pass = true; 403 ret = 0; 404 } 405 406 if (ret == DROP_THIS_NODE) { 407 six_unlock_read(&cur->c.lock); 408 bch2_btree_node_evict(trans, cur_k.k); 409 ret = bch2_journal_key_delete(c, b->c.btree_id, 410 b->c.level, cur_k.k->k.p); 411 cur = NULL; 412 if (ret) 413 break; 414 continue; 415 } 416 417 if (prev) 418 six_unlock_read(&prev->c.lock); 419 prev = NULL; 420 421 if (ret == DROP_PREV_NODE) { 422 bch_info(c, "dropped prev node"); 423 bch2_btree_node_evict(trans, prev_k.k); 424 ret = bch2_journal_key_delete(c, b->c.btree_id, 425 b->c.level, prev_k.k->k.p); 426 if (ret) 427 break; 428 429 bch2_btree_and_journal_iter_exit(&iter); 430 goto again; 431 } else if (ret) 432 break; 433 434 prev = cur; 435 cur = NULL; 436 bch2_bkey_buf_copy(&prev_k, c, cur_k.k); 437 } 438 439 if (!ret && !IS_ERR_OR_NULL(prev)) { 440 BUG_ON(cur); 441 ret = btree_repair_node_end(c, b, prev, pulled_from_scan); 442 if (ret == DID_FILL_FROM_SCAN) { 443 new_pass = true; 444 ret = 0; 445 } 446 } 447 448 if (!IS_ERR_OR_NULL(prev)) 449 six_unlock_read(&prev->c.lock); 450 prev = NULL; 451 if (!IS_ERR_OR_NULL(cur)) 452 six_unlock_read(&cur->c.lock); 453 cur = NULL; 454 455 if (ret) 456 goto err; 457 458 bch2_btree_and_journal_iter_exit(&iter); 459 460 if (new_pass) 461 goto again; 462 463 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 464 iter.prefetch = true; 465 466 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 467 bch2_bkey_buf_reassemble(&cur_k, c, k); 468 bch2_btree_and_journal_iter_advance(&iter); 469 470 cur = bch2_btree_node_get_noiter(trans, cur_k.k, 471 b->c.btree_id, b->c.level - 1, 472 false); 473 ret = PTR_ERR_OR_ZERO(cur); 474 475 bch_err_msg(c, ret, "getting btree node"); 476 if (ret) 477 goto err; 478 479 ret = bch2_btree_repair_topology_recurse(trans, cur, pulled_from_scan); 480 six_unlock_read(&cur->c.lock); 481 cur = NULL; 482 483 if (ret == DROP_THIS_NODE) { 484 bch2_btree_node_evict(trans, cur_k.k); 485 ret = bch2_journal_key_delete(c, b->c.btree_id, 486 b->c.level, cur_k.k->k.p); 487 new_pass = true; 488 } 489 490 if (ret) 491 goto err; 492 493 have_child = true; 494 } 495 496 printbuf_reset(&buf); 497 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); 498 499 if (mustfix_fsck_err_on(!have_child, c, 500 btree_node_topology_interior_node_empty, 501 "empty interior btree node at btree %s level %u\n" 502 " %s", 503 bch2_btree_id_str(b->c.btree_id), 504 b->c.level, buf.buf)) 505 ret = DROP_THIS_NODE; 506 err: 507 fsck_err: 508 if (!IS_ERR_OR_NULL(prev)) 509 six_unlock_read(&prev->c.lock); 510 if (!IS_ERR_OR_NULL(cur)) 511 six_unlock_read(&cur->c.lock); 512 513 bch2_btree_and_journal_iter_exit(&iter); 514 515 if (!ret && new_pass) 516 goto again; 517 518 BUG_ON(!ret && bch2_btree_node_check_topology(trans, b)); 519 520 bch2_bkey_buf_exit(&prev_k, c); 521 bch2_bkey_buf_exit(&cur_k, c); 522 printbuf_exit(&buf); 523 return ret; 524 } 525 526 int bch2_check_topology(struct bch_fs *c) 527 { 528 struct btree_trans *trans = bch2_trans_get(c); 529 struct bpos pulled_from_scan = POS_MIN; 530 int ret = 0; 531 532 for (unsigned i = 0; i < btree_id_nr_alive(c) && !ret; i++) { 533 struct btree_root *r = bch2_btree_id_root(c, i); 534 bool reconstructed_root = false; 535 536 if (r->error) { 537 ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes); 538 if (ret) 539 break; 540 reconstruct_root: 541 bch_info(c, "btree root %s unreadable, must recover from scan", bch2_btree_id_str(i)); 542 543 r->alive = false; 544 r->error = 0; 545 546 if (!bch2_btree_has_scanned_nodes(c, i)) { 547 mustfix_fsck_err(c, btree_root_unreadable_and_scan_found_nothing, 548 "no nodes found for btree %s, continue?", bch2_btree_id_str(i)); 549 bch2_btree_root_alloc_fake(c, i, 0); 550 } else { 551 bch2_btree_root_alloc_fake(c, i, 1); 552 bch2_shoot_down_journal_keys(c, i, 1, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX); 553 ret = bch2_get_scanned_nodes(c, i, 0, POS_MIN, SPOS_MAX); 554 if (ret) 555 break; 556 } 557 558 reconstructed_root = true; 559 } 560 561 struct btree *b = r->b; 562 563 btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_read); 564 ret = bch2_btree_repair_topology_recurse(trans, b, &pulled_from_scan); 565 six_unlock_read(&b->c.lock); 566 567 if (ret == DROP_THIS_NODE) { 568 bch2_btree_node_hash_remove(&c->btree_cache, b); 569 mutex_lock(&c->btree_cache.lock); 570 list_move(&b->list, &c->btree_cache.freeable); 571 mutex_unlock(&c->btree_cache.lock); 572 573 r->b = NULL; 574 575 if (!reconstructed_root) 576 goto reconstruct_root; 577 578 bch_err(c, "empty btree root %s", bch2_btree_id_str(i)); 579 bch2_btree_root_alloc_fake(c, i, 0); 580 r->alive = false; 581 ret = 0; 582 } 583 } 584 fsck_err: 585 bch2_trans_put(trans); 586 return ret; 587 } 588 589 static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id, 590 unsigned level, bool is_root, 591 struct bkey_s_c *k) 592 { 593 struct bch_fs *c = trans->c; 594 struct bkey_ptrs_c ptrs_c = bch2_bkey_ptrs_c(*k); 595 const union bch_extent_entry *entry_c; 596 struct extent_ptr_decoded p = { 0 }; 597 bool do_update = false; 598 struct printbuf buf = PRINTBUF; 599 int ret = 0; 600 601 /* 602 * XXX 603 * use check_bucket_ref here 604 */ 605 bkey_for_each_ptr_decode(k->k, ptrs_c, p, entry_c) { 606 struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); 607 struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr); 608 enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, p, entry_c); 609 610 if (fsck_err_on(!g->gen_valid, 611 c, ptr_to_missing_alloc_key, 612 "bucket %u:%zu data type %s ptr gen %u missing in alloc btree\n" 613 "while marking %s", 614 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 615 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 616 p.ptr.gen, 617 (printbuf_reset(&buf), 618 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) { 619 if (!p.ptr.cached) { 620 g->gen_valid = true; 621 g->gen = p.ptr.gen; 622 } else { 623 do_update = true; 624 } 625 } 626 627 if (fsck_err_on(gen_cmp(p.ptr.gen, g->gen) > 0, 628 c, ptr_gen_newer_than_bucket_gen, 629 "bucket %u:%zu data type %s ptr gen in the future: %u > %u\n" 630 "while marking %s", 631 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 632 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 633 p.ptr.gen, g->gen, 634 (printbuf_reset(&buf), 635 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) { 636 if (!p.ptr.cached) { 637 g->gen_valid = true; 638 g->gen = p.ptr.gen; 639 g->data_type = 0; 640 g->dirty_sectors = 0; 641 g->cached_sectors = 0; 642 set_bit(BCH_FS_need_another_gc, &c->flags); 643 } else { 644 do_update = true; 645 } 646 } 647 648 if (fsck_err_on(gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX, 649 c, ptr_gen_newer_than_bucket_gen, 650 "bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n" 651 "while marking %s", 652 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), g->gen, 653 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 654 p.ptr.gen, 655 (printbuf_reset(&buf), 656 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 657 do_update = true; 658 659 if (fsck_err_on(!p.ptr.cached && gen_cmp(p.ptr.gen, g->gen) < 0, 660 c, stale_dirty_ptr, 661 "bucket %u:%zu data type %s stale dirty ptr: %u < %u\n" 662 "while marking %s", 663 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 664 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 665 p.ptr.gen, g->gen, 666 (printbuf_reset(&buf), 667 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 668 do_update = true; 669 670 if (data_type != BCH_DATA_btree && p.ptr.gen != g->gen) 671 continue; 672 673 if (fsck_err_on(bucket_data_type(g->data_type) && 674 bucket_data_type(g->data_type) != 675 bucket_data_type(data_type), c, 676 ptr_bucket_data_type_mismatch, 677 "bucket %u:%zu different types of data in same bucket: %s, %s\n" 678 "while marking %s", 679 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 680 bch2_data_type_str(g->data_type), 681 bch2_data_type_str(data_type), 682 (printbuf_reset(&buf), 683 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) { 684 if (data_type == BCH_DATA_btree) { 685 g->data_type = data_type; 686 set_bit(BCH_FS_need_another_gc, &c->flags); 687 } else { 688 do_update = true; 689 } 690 } 691 692 if (p.has_ec) { 693 struct gc_stripe *m = genradix_ptr(&c->gc_stripes, p.ec.idx); 694 695 if (fsck_err_on(!m || !m->alive, c, 696 ptr_to_missing_stripe, 697 "pointer to nonexistent stripe %llu\n" 698 "while marking %s", 699 (u64) p.ec.idx, 700 (printbuf_reset(&buf), 701 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 702 do_update = true; 703 704 if (fsck_err_on(m && m->alive && !bch2_ptr_matches_stripe_m(m, p), c, 705 ptr_to_incorrect_stripe, 706 "pointer does not match stripe %llu\n" 707 "while marking %s", 708 (u64) p.ec.idx, 709 (printbuf_reset(&buf), 710 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 711 do_update = true; 712 } 713 } 714 715 if (do_update) { 716 if (is_root) { 717 bch_err(c, "cannot update btree roots yet"); 718 ret = -EINVAL; 719 goto err; 720 } 721 722 struct bkey_i *new = kmalloc(bkey_bytes(k->k), GFP_KERNEL); 723 if (!new) { 724 ret = -BCH_ERR_ENOMEM_gc_repair_key; 725 bch_err_msg(c, ret, "allocating new key"); 726 goto err; 727 } 728 729 bkey_reassemble(new, *k); 730 731 if (level) { 732 /* 733 * We don't want to drop btree node pointers - if the 734 * btree node isn't there anymore, the read path will 735 * sort it out: 736 */ 737 struct bkey_ptrs ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); 738 bkey_for_each_ptr(ptrs, ptr) { 739 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 740 struct bucket *g = PTR_GC_BUCKET(ca, ptr); 741 742 ptr->gen = g->gen; 743 } 744 } else { 745 struct bkey_ptrs ptrs; 746 union bch_extent_entry *entry; 747 restart_drop_ptrs: 748 ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); 749 bkey_for_each_ptr_decode(bkey_i_to_s(new).k, ptrs, p, entry) { 750 struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); 751 struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr); 752 enum bch_data_type data_type = bch2_bkey_ptr_data_type(bkey_i_to_s_c(new), p, entry); 753 754 if ((p.ptr.cached && 755 (!g->gen_valid || gen_cmp(p.ptr.gen, g->gen) > 0)) || 756 (!p.ptr.cached && 757 gen_cmp(p.ptr.gen, g->gen) < 0) || 758 gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX || 759 (g->data_type && 760 g->data_type != data_type)) { 761 bch2_bkey_drop_ptr(bkey_i_to_s(new), &entry->ptr); 762 goto restart_drop_ptrs; 763 } 764 } 765 again: 766 ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); 767 bkey_extent_entry_for_each(ptrs, entry) { 768 if (extent_entry_type(entry) == BCH_EXTENT_ENTRY_stripe_ptr) { 769 struct gc_stripe *m = genradix_ptr(&c->gc_stripes, 770 entry->stripe_ptr.idx); 771 union bch_extent_entry *next_ptr; 772 773 bkey_extent_entry_for_each_from(ptrs, next_ptr, entry) 774 if (extent_entry_type(next_ptr) == BCH_EXTENT_ENTRY_ptr) 775 goto found; 776 next_ptr = NULL; 777 found: 778 if (!next_ptr) { 779 bch_err(c, "aieee, found stripe ptr with no data ptr"); 780 continue; 781 } 782 783 if (!m || !m->alive || 784 !__bch2_ptr_matches_stripe(&m->ptrs[entry->stripe_ptr.block], 785 &next_ptr->ptr, 786 m->sectors)) { 787 bch2_bkey_extent_entry_drop(new, entry); 788 goto again; 789 } 790 } 791 } 792 } 793 794 if (level) 795 bch2_btree_node_update_key_early(trans, btree_id, level - 1, *k, new); 796 797 if (0) { 798 printbuf_reset(&buf); 799 bch2_bkey_val_to_text(&buf, c, *k); 800 bch_info(c, "updated %s", buf.buf); 801 802 printbuf_reset(&buf); 803 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(new)); 804 bch_info(c, "new key %s", buf.buf); 805 } 806 807 ret = bch2_journal_key_insert_take(c, btree_id, level, new); 808 if (ret) { 809 kfree(new); 810 goto err; 811 } 812 813 *k = bkey_i_to_s_c(new); 814 } 815 err: 816 fsck_err: 817 printbuf_exit(&buf); 818 return ret; 819 } 820 821 /* marking of btree keys/nodes: */ 822 823 static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id, 824 unsigned level, bool is_root, 825 struct bkey_s_c *k, 826 bool initial) 827 { 828 struct bch_fs *c = trans->c; 829 struct bkey deleted = KEY(0, 0, 0); 830 struct bkey_s_c old = (struct bkey_s_c) { &deleted, NULL }; 831 int ret = 0; 832 833 deleted.p = k->k->p; 834 835 if (initial) { 836 BUG_ON(bch2_journal_seq_verify && 837 k->k->version.lo > atomic64_read(&c->journal.seq)); 838 839 if (fsck_err_on(k->k->version.lo > atomic64_read(&c->key_version), c, 840 bkey_version_in_future, 841 "key version number higher than recorded: %llu > %llu", 842 k->k->version.lo, 843 atomic64_read(&c->key_version))) 844 atomic64_set(&c->key_version, k->k->version.lo); 845 } 846 847 ret = bch2_check_fix_ptrs(trans, btree_id, level, is_root, k); 848 if (ret) 849 goto err; 850 851 ret = commit_do(trans, NULL, NULL, 0, 852 bch2_key_trigger(trans, btree_id, level, old, 853 unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC)); 854 fsck_err: 855 err: 856 bch_err_fn(c, ret); 857 return ret; 858 } 859 860 static int btree_gc_mark_node(struct btree_trans *trans, struct btree *b, bool initial) 861 { 862 struct btree_node_iter iter; 863 struct bkey unpacked; 864 struct bkey_s_c k; 865 int ret = 0; 866 867 ret = bch2_btree_node_check_topology(trans, b); 868 if (ret) 869 return ret; 870 871 if (!btree_node_type_needs_gc(btree_node_type(b))) 872 return 0; 873 874 bch2_btree_node_iter_init_from_start(&iter, b); 875 876 while ((k = bch2_btree_node_iter_peek_unpack(&iter, b, &unpacked)).k) { 877 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, false, 878 &k, initial); 879 if (ret) 880 return ret; 881 882 bch2_btree_node_iter_advance(&iter, b); 883 } 884 885 return 0; 886 } 887 888 static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id, 889 bool initial, bool metadata_only) 890 { 891 struct bch_fs *c = trans->c; 892 struct btree_iter iter; 893 struct btree *b; 894 unsigned depth = metadata_only ? 1 : 0; 895 int ret = 0; 896 897 gc_pos_set(c, gc_pos_btree(btree_id, POS_MIN, 0)); 898 899 __for_each_btree_node(trans, iter, btree_id, POS_MIN, 900 0, depth, BTREE_ITER_PREFETCH, b, ret) { 901 bch2_verify_btree_nr_keys(b); 902 903 gc_pos_set(c, gc_pos_btree_node(b)); 904 905 ret = btree_gc_mark_node(trans, b, initial); 906 if (ret) 907 break; 908 } 909 bch2_trans_iter_exit(trans, &iter); 910 911 if (ret) 912 return ret; 913 914 mutex_lock(&c->btree_root_lock); 915 b = bch2_btree_id_root(c, btree_id)->b; 916 if (!btree_node_fake(b)) { 917 struct bkey_s_c k = bkey_i_to_s_c(&b->key); 918 919 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1, 920 true, &k, initial); 921 } 922 gc_pos_set(c, gc_pos_btree_root(b->c.btree_id)); 923 mutex_unlock(&c->btree_root_lock); 924 925 return ret; 926 } 927 928 static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b, 929 unsigned target_depth) 930 { 931 struct bch_fs *c = trans->c; 932 struct btree_and_journal_iter iter; 933 struct bkey_s_c k; 934 struct bkey_buf cur; 935 struct printbuf buf = PRINTBUF; 936 int ret = 0; 937 938 ret = bch2_btree_node_check_topology(trans, b); 939 if (ret) 940 return ret; 941 942 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 943 bch2_bkey_buf_init(&cur); 944 945 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 946 BUG_ON(bpos_lt(k.k->p, b->data->min_key)); 947 BUG_ON(bpos_gt(k.k->p, b->data->max_key)); 948 949 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, 950 false, &k, true); 951 if (ret) 952 goto fsck_err; 953 954 bch2_btree_and_journal_iter_advance(&iter); 955 } 956 957 if (b->c.level > target_depth) { 958 bch2_btree_and_journal_iter_exit(&iter); 959 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 960 iter.prefetch = true; 961 962 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 963 struct btree *child; 964 965 bch2_bkey_buf_reassemble(&cur, c, k); 966 bch2_btree_and_journal_iter_advance(&iter); 967 968 child = bch2_btree_node_get_noiter(trans, cur.k, 969 b->c.btree_id, b->c.level - 1, 970 false); 971 ret = PTR_ERR_OR_ZERO(child); 972 973 if (bch2_err_matches(ret, EIO)) { 974 bch2_topology_error(c); 975 976 if (__fsck_err(c, 977 FSCK_CAN_FIX| 978 FSCK_CAN_IGNORE| 979 FSCK_NO_RATELIMIT, 980 btree_node_read_error, 981 "Unreadable btree node at btree %s level %u:\n" 982 " %s", 983 bch2_btree_id_str(b->c.btree_id), 984 b->c.level - 1, 985 (printbuf_reset(&buf), 986 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur.k)), buf.buf)) && 987 should_restart_for_topology_repair(c)) { 988 bch_info(c, "Halting mark and sweep to start topology repair pass"); 989 ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); 990 goto fsck_err; 991 } else { 992 /* Continue marking when opted to not 993 * fix the error: */ 994 ret = 0; 995 set_bit(BCH_FS_initial_gc_unfixed, &c->flags); 996 continue; 997 } 998 } else if (ret) { 999 bch_err_msg(c, ret, "getting btree node"); 1000 break; 1001 } 1002 1003 ret = bch2_gc_btree_init_recurse(trans, child, 1004 target_depth); 1005 six_unlock_read(&child->c.lock); 1006 1007 if (ret) 1008 break; 1009 } 1010 } 1011 fsck_err: 1012 bch2_bkey_buf_exit(&cur, c); 1013 bch2_btree_and_journal_iter_exit(&iter); 1014 printbuf_exit(&buf); 1015 return ret; 1016 } 1017 1018 static int bch2_gc_btree_init(struct btree_trans *trans, 1019 enum btree_id btree_id, 1020 bool metadata_only) 1021 { 1022 struct bch_fs *c = trans->c; 1023 struct btree *b; 1024 unsigned target_depth = metadata_only ? 1 : 0; 1025 struct printbuf buf = PRINTBUF; 1026 int ret = 0; 1027 1028 b = bch2_btree_id_root(c, btree_id)->b; 1029 1030 six_lock_read(&b->c.lock, NULL, NULL); 1031 printbuf_reset(&buf); 1032 bch2_bpos_to_text(&buf, b->data->min_key); 1033 if (mustfix_fsck_err_on(!bpos_eq(b->data->min_key, POS_MIN), c, 1034 btree_root_bad_min_key, 1035 "btree root with incorrect min_key: %s", buf.buf)) { 1036 bch_err(c, "repair unimplemented"); 1037 ret = -BCH_ERR_fsck_repair_unimplemented; 1038 goto fsck_err; 1039 } 1040 1041 printbuf_reset(&buf); 1042 bch2_bpos_to_text(&buf, b->data->max_key); 1043 if (mustfix_fsck_err_on(!bpos_eq(b->data->max_key, SPOS_MAX), c, 1044 btree_root_bad_max_key, 1045 "btree root with incorrect max_key: %s", buf.buf)) { 1046 bch_err(c, "repair unimplemented"); 1047 ret = -BCH_ERR_fsck_repair_unimplemented; 1048 goto fsck_err; 1049 } 1050 1051 if (b->c.level >= target_depth) 1052 ret = bch2_gc_btree_init_recurse(trans, b, target_depth); 1053 1054 if (!ret) { 1055 struct bkey_s_c k = bkey_i_to_s_c(&b->key); 1056 1057 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1, true, 1058 &k, true); 1059 } 1060 fsck_err: 1061 six_unlock_read(&b->c.lock); 1062 1063 bch_err_fn(c, ret); 1064 printbuf_exit(&buf); 1065 return ret; 1066 } 1067 1068 static inline int btree_id_gc_phase_cmp(enum btree_id l, enum btree_id r) 1069 { 1070 return (int) btree_id_to_gc_phase(l) - 1071 (int) btree_id_to_gc_phase(r); 1072 } 1073 1074 static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only) 1075 { 1076 struct btree_trans *trans = bch2_trans_get(c); 1077 enum btree_id ids[BTREE_ID_NR]; 1078 unsigned i; 1079 int ret = 0; 1080 1081 for (i = 0; i < BTREE_ID_NR; i++) 1082 ids[i] = i; 1083 bubble_sort(ids, BTREE_ID_NR, btree_id_gc_phase_cmp); 1084 1085 for (i = 0; i < BTREE_ID_NR && !ret; i++) 1086 ret = initial 1087 ? bch2_gc_btree_init(trans, ids[i], metadata_only) 1088 : bch2_gc_btree(trans, ids[i], initial, metadata_only); 1089 1090 for (i = BTREE_ID_NR; i < btree_id_nr_alive(c) && !ret; i++) { 1091 if (!bch2_btree_id_root(c, i)->alive) 1092 continue; 1093 1094 ret = initial 1095 ? bch2_gc_btree_init(trans, i, metadata_only) 1096 : bch2_gc_btree(trans, i, initial, metadata_only); 1097 } 1098 1099 bch2_trans_put(trans); 1100 bch_err_fn(c, ret); 1101 return ret; 1102 } 1103 1104 static void mark_metadata_sectors(struct bch_fs *c, struct bch_dev *ca, 1105 u64 start, u64 end, 1106 enum bch_data_type type, 1107 unsigned flags) 1108 { 1109 u64 b = sector_to_bucket(ca, start); 1110 1111 do { 1112 unsigned sectors = 1113 min_t(u64, bucket_to_sector(ca, b + 1), end) - start; 1114 1115 bch2_mark_metadata_bucket(c, ca, b, type, sectors, 1116 gc_phase(GC_PHASE_SB), flags); 1117 b++; 1118 start += sectors; 1119 } while (start < end); 1120 } 1121 1122 static void bch2_mark_dev_superblock(struct bch_fs *c, struct bch_dev *ca, 1123 unsigned flags) 1124 { 1125 struct bch_sb_layout *layout = &ca->disk_sb.sb->layout; 1126 unsigned i; 1127 u64 b; 1128 1129 for (i = 0; i < layout->nr_superblocks; i++) { 1130 u64 offset = le64_to_cpu(layout->sb_offset[i]); 1131 1132 if (offset == BCH_SB_SECTOR) 1133 mark_metadata_sectors(c, ca, 0, BCH_SB_SECTOR, 1134 BCH_DATA_sb, flags); 1135 1136 mark_metadata_sectors(c, ca, offset, 1137 offset + (1 << layout->sb_max_size_bits), 1138 BCH_DATA_sb, flags); 1139 } 1140 1141 for (i = 0; i < ca->journal.nr; i++) { 1142 b = ca->journal.buckets[i]; 1143 bch2_mark_metadata_bucket(c, ca, b, BCH_DATA_journal, 1144 ca->mi.bucket_size, 1145 gc_phase(GC_PHASE_SB), flags); 1146 } 1147 } 1148 1149 static void bch2_mark_superblocks(struct bch_fs *c) 1150 { 1151 mutex_lock(&c->sb_lock); 1152 gc_pos_set(c, gc_phase(GC_PHASE_SB)); 1153 1154 for_each_online_member(c, ca) 1155 bch2_mark_dev_superblock(c, ca, BTREE_TRIGGER_GC); 1156 mutex_unlock(&c->sb_lock); 1157 } 1158 1159 #if 0 1160 /* Also see bch2_pending_btree_node_free_insert_done() */ 1161 static void bch2_mark_pending_btree_node_frees(struct bch_fs *c) 1162 { 1163 struct btree_update *as; 1164 struct pending_btree_node_free *d; 1165 1166 mutex_lock(&c->btree_interior_update_lock); 1167 gc_pos_set(c, gc_phase(GC_PHASE_PENDING_DELETE)); 1168 1169 for_each_pending_btree_node_free(c, as, d) 1170 if (d->index_update_done) 1171 bch2_mark_key(c, bkey_i_to_s_c(&d->key), BTREE_TRIGGER_GC); 1172 1173 mutex_unlock(&c->btree_interior_update_lock); 1174 } 1175 #endif 1176 1177 static void bch2_gc_free(struct bch_fs *c) 1178 { 1179 genradix_free(&c->reflink_gc_table); 1180 genradix_free(&c->gc_stripes); 1181 1182 for_each_member_device(c, ca) { 1183 kvfree(rcu_dereference_protected(ca->buckets_gc, 1)); 1184 ca->buckets_gc = NULL; 1185 1186 free_percpu(ca->usage_gc); 1187 ca->usage_gc = NULL; 1188 } 1189 1190 free_percpu(c->usage_gc); 1191 c->usage_gc = NULL; 1192 } 1193 1194 static int bch2_gc_done(struct bch_fs *c, 1195 bool initial, bool metadata_only) 1196 { 1197 struct bch_dev *ca = NULL; 1198 struct printbuf buf = PRINTBUF; 1199 bool verify = !metadata_only && 1200 !c->opts.reconstruct_alloc && 1201 (!initial || (c->sb.compat & (1ULL << BCH_COMPAT_alloc_info))); 1202 unsigned i; 1203 int ret = 0; 1204 1205 percpu_down_write(&c->mark_lock); 1206 1207 #define copy_field(_err, _f, _msg, ...) \ 1208 if (dst->_f != src->_f && \ 1209 (!verify || \ 1210 fsck_err(c, _err, _msg ": got %llu, should be %llu" \ 1211 , ##__VA_ARGS__, dst->_f, src->_f))) \ 1212 dst->_f = src->_f 1213 #define copy_dev_field(_err, _f, _msg, ...) \ 1214 copy_field(_err, _f, "dev %u has wrong " _msg, ca->dev_idx, ##__VA_ARGS__) 1215 #define copy_fs_field(_err, _f, _msg, ...) \ 1216 copy_field(_err, _f, "fs has wrong " _msg, ##__VA_ARGS__) 1217 1218 for (i = 0; i < ARRAY_SIZE(c->usage); i++) 1219 bch2_fs_usage_acc_to_base(c, i); 1220 1221 __for_each_member_device(c, ca) { 1222 struct bch_dev_usage *dst = ca->usage_base; 1223 struct bch_dev_usage *src = (void *) 1224 bch2_acc_percpu_u64s((u64 __percpu *) ca->usage_gc, 1225 dev_usage_u64s()); 1226 1227 for (i = 0; i < BCH_DATA_NR; i++) { 1228 copy_dev_field(dev_usage_buckets_wrong, 1229 d[i].buckets, "%s buckets", bch2_data_type_str(i)); 1230 copy_dev_field(dev_usage_sectors_wrong, 1231 d[i].sectors, "%s sectors", bch2_data_type_str(i)); 1232 copy_dev_field(dev_usage_fragmented_wrong, 1233 d[i].fragmented, "%s fragmented", bch2_data_type_str(i)); 1234 } 1235 } 1236 1237 { 1238 unsigned nr = fs_usage_u64s(c); 1239 struct bch_fs_usage *dst = c->usage_base; 1240 struct bch_fs_usage *src = (void *) 1241 bch2_acc_percpu_u64s((u64 __percpu *) c->usage_gc, nr); 1242 1243 copy_fs_field(fs_usage_hidden_wrong, 1244 b.hidden, "hidden"); 1245 copy_fs_field(fs_usage_btree_wrong, 1246 b.btree, "btree"); 1247 1248 if (!metadata_only) { 1249 copy_fs_field(fs_usage_data_wrong, 1250 b.data, "data"); 1251 copy_fs_field(fs_usage_cached_wrong, 1252 b.cached, "cached"); 1253 copy_fs_field(fs_usage_reserved_wrong, 1254 b.reserved, "reserved"); 1255 copy_fs_field(fs_usage_nr_inodes_wrong, 1256 b.nr_inodes,"nr_inodes"); 1257 1258 for (i = 0; i < BCH_REPLICAS_MAX; i++) 1259 copy_fs_field(fs_usage_persistent_reserved_wrong, 1260 persistent_reserved[i], 1261 "persistent_reserved[%i]", i); 1262 } 1263 1264 for (i = 0; i < c->replicas.nr; i++) { 1265 struct bch_replicas_entry_v1 *e = 1266 cpu_replicas_entry(&c->replicas, i); 1267 1268 if (metadata_only && 1269 (e->data_type == BCH_DATA_user || 1270 e->data_type == BCH_DATA_cached)) 1271 continue; 1272 1273 printbuf_reset(&buf); 1274 bch2_replicas_entry_to_text(&buf, e); 1275 1276 copy_fs_field(fs_usage_replicas_wrong, 1277 replicas[i], "%s", buf.buf); 1278 } 1279 } 1280 1281 #undef copy_fs_field 1282 #undef copy_dev_field 1283 #undef copy_stripe_field 1284 #undef copy_field 1285 fsck_err: 1286 if (ca) 1287 percpu_ref_put(&ca->ref); 1288 bch_err_fn(c, ret); 1289 1290 percpu_up_write(&c->mark_lock); 1291 printbuf_exit(&buf); 1292 return ret; 1293 } 1294 1295 static int bch2_gc_start(struct bch_fs *c) 1296 { 1297 BUG_ON(c->usage_gc); 1298 1299 c->usage_gc = __alloc_percpu_gfp(fs_usage_u64s(c) * sizeof(u64), 1300 sizeof(u64), GFP_KERNEL); 1301 if (!c->usage_gc) { 1302 bch_err(c, "error allocating c->usage_gc"); 1303 return -BCH_ERR_ENOMEM_gc_start; 1304 } 1305 1306 for_each_member_device(c, ca) { 1307 BUG_ON(ca->usage_gc); 1308 1309 ca->usage_gc = alloc_percpu(struct bch_dev_usage); 1310 if (!ca->usage_gc) { 1311 bch_err(c, "error allocating ca->usage_gc"); 1312 percpu_ref_put(&ca->ref); 1313 return -BCH_ERR_ENOMEM_gc_start; 1314 } 1315 1316 this_cpu_write(ca->usage_gc->d[BCH_DATA_free].buckets, 1317 ca->mi.nbuckets - ca->mi.first_bucket); 1318 } 1319 1320 return 0; 1321 } 1322 1323 static int bch2_gc_reset(struct bch_fs *c) 1324 { 1325 for_each_member_device(c, ca) { 1326 free_percpu(ca->usage_gc); 1327 ca->usage_gc = NULL; 1328 } 1329 1330 free_percpu(c->usage_gc); 1331 c->usage_gc = NULL; 1332 1333 return bch2_gc_start(c); 1334 } 1335 1336 /* returns true if not equal */ 1337 static inline bool bch2_alloc_v4_cmp(struct bch_alloc_v4 l, 1338 struct bch_alloc_v4 r) 1339 { 1340 return l.gen != r.gen || 1341 l.oldest_gen != r.oldest_gen || 1342 l.data_type != r.data_type || 1343 l.dirty_sectors != r.dirty_sectors || 1344 l.cached_sectors != r.cached_sectors || 1345 l.stripe_redundancy != r.stripe_redundancy || 1346 l.stripe != r.stripe; 1347 } 1348 1349 static int bch2_alloc_write_key(struct btree_trans *trans, 1350 struct btree_iter *iter, 1351 struct bkey_s_c k, 1352 bool metadata_only) 1353 { 1354 struct bch_fs *c = trans->c; 1355 struct bch_dev *ca = bch_dev_bkey_exists(c, iter->pos.inode); 1356 struct bucket old_gc, gc, *b; 1357 struct bkey_i_alloc_v4 *a; 1358 struct bch_alloc_v4 old_convert, new; 1359 const struct bch_alloc_v4 *old; 1360 int ret; 1361 1362 old = bch2_alloc_to_v4(k, &old_convert); 1363 new = *old; 1364 1365 percpu_down_read(&c->mark_lock); 1366 b = gc_bucket(ca, iter->pos.offset); 1367 old_gc = *b; 1368 1369 if ((old->data_type == BCH_DATA_sb || 1370 old->data_type == BCH_DATA_journal) && 1371 !bch2_dev_is_online(ca)) { 1372 b->data_type = old->data_type; 1373 b->dirty_sectors = old->dirty_sectors; 1374 } 1375 1376 /* 1377 * b->data_type doesn't yet include need_discard & need_gc_gen states - 1378 * fix that here: 1379 */ 1380 b->data_type = __alloc_data_type(b->dirty_sectors, 1381 b->cached_sectors, 1382 b->stripe, 1383 *old, 1384 b->data_type); 1385 gc = *b; 1386 1387 if (gc.data_type != old_gc.data_type || 1388 gc.dirty_sectors != old_gc.dirty_sectors) 1389 bch2_dev_usage_update_m(c, ca, &old_gc, &gc); 1390 percpu_up_read(&c->mark_lock); 1391 1392 if (metadata_only && 1393 gc.data_type != BCH_DATA_sb && 1394 gc.data_type != BCH_DATA_journal && 1395 gc.data_type != BCH_DATA_btree) 1396 return 0; 1397 1398 if (gen_after(old->gen, gc.gen)) 1399 return 0; 1400 1401 if (fsck_err_on(new.data_type != gc.data_type, c, 1402 alloc_key_data_type_wrong, 1403 "bucket %llu:%llu gen %u has wrong data_type" 1404 ": got %s, should be %s", 1405 iter->pos.inode, iter->pos.offset, 1406 gc.gen, 1407 bch2_data_type_str(new.data_type), 1408 bch2_data_type_str(gc.data_type))) 1409 new.data_type = gc.data_type; 1410 1411 #define copy_bucket_field(_errtype, _f) \ 1412 if (fsck_err_on(new._f != gc._f, c, _errtype, \ 1413 "bucket %llu:%llu gen %u data type %s has wrong " #_f \ 1414 ": got %u, should be %u", \ 1415 iter->pos.inode, iter->pos.offset, \ 1416 gc.gen, \ 1417 bch2_data_type_str(gc.data_type), \ 1418 new._f, gc._f)) \ 1419 new._f = gc._f; \ 1420 1421 copy_bucket_field(alloc_key_gen_wrong, 1422 gen); 1423 copy_bucket_field(alloc_key_dirty_sectors_wrong, 1424 dirty_sectors); 1425 copy_bucket_field(alloc_key_cached_sectors_wrong, 1426 cached_sectors); 1427 copy_bucket_field(alloc_key_stripe_wrong, 1428 stripe); 1429 copy_bucket_field(alloc_key_stripe_redundancy_wrong, 1430 stripe_redundancy); 1431 #undef copy_bucket_field 1432 1433 if (!bch2_alloc_v4_cmp(*old, new)) 1434 return 0; 1435 1436 a = bch2_alloc_to_v4_mut(trans, k); 1437 ret = PTR_ERR_OR_ZERO(a); 1438 if (ret) 1439 return ret; 1440 1441 a->v = new; 1442 1443 /* 1444 * The trigger normally makes sure this is set, but we're not running 1445 * triggers: 1446 */ 1447 if (a->v.data_type == BCH_DATA_cached && !a->v.io_time[READ]) 1448 a->v.io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now)); 1449 1450 ret = bch2_trans_update(trans, iter, &a->k_i, BTREE_TRIGGER_NORUN); 1451 fsck_err: 1452 return ret; 1453 } 1454 1455 static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only) 1456 { 1457 int ret = 0; 1458 1459 for_each_member_device(c, ca) { 1460 ret = bch2_trans_run(c, 1461 for_each_btree_key_upto_commit(trans, iter, BTREE_ID_alloc, 1462 POS(ca->dev_idx, ca->mi.first_bucket), 1463 POS(ca->dev_idx, ca->mi.nbuckets - 1), 1464 BTREE_ITER_SLOTS|BTREE_ITER_PREFETCH, k, 1465 NULL, NULL, BCH_TRANS_COMMIT_lazy_rw, 1466 bch2_alloc_write_key(trans, &iter, k, metadata_only))); 1467 if (ret) { 1468 percpu_ref_put(&ca->ref); 1469 break; 1470 } 1471 } 1472 1473 bch_err_fn(c, ret); 1474 return ret; 1475 } 1476 1477 static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only) 1478 { 1479 for_each_member_device(c, ca) { 1480 struct bucket_array *buckets = kvmalloc(sizeof(struct bucket_array) + 1481 ca->mi.nbuckets * sizeof(struct bucket), 1482 GFP_KERNEL|__GFP_ZERO); 1483 if (!buckets) { 1484 percpu_ref_put(&ca->ref); 1485 bch_err(c, "error allocating ca->buckets[gc]"); 1486 return -BCH_ERR_ENOMEM_gc_alloc_start; 1487 } 1488 1489 buckets->first_bucket = ca->mi.first_bucket; 1490 buckets->nbuckets = ca->mi.nbuckets; 1491 rcu_assign_pointer(ca->buckets_gc, buckets); 1492 } 1493 1494 int ret = bch2_trans_run(c, 1495 for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN, 1496 BTREE_ITER_PREFETCH, k, ({ 1497 struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode); 1498 struct bucket *g = gc_bucket(ca, k.k->p.offset); 1499 1500 struct bch_alloc_v4 a_convert; 1501 const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert); 1502 1503 g->gen_valid = 1; 1504 g->gen = a->gen; 1505 1506 if (metadata_only && 1507 (a->data_type == BCH_DATA_user || 1508 a->data_type == BCH_DATA_cached || 1509 a->data_type == BCH_DATA_parity)) { 1510 g->data_type = a->data_type; 1511 g->dirty_sectors = a->dirty_sectors; 1512 g->cached_sectors = a->cached_sectors; 1513 g->stripe = a->stripe; 1514 g->stripe_redundancy = a->stripe_redundancy; 1515 } 1516 1517 0; 1518 }))); 1519 bch_err_fn(c, ret); 1520 return ret; 1521 } 1522 1523 static void bch2_gc_alloc_reset(struct bch_fs *c, bool metadata_only) 1524 { 1525 for_each_member_device(c, ca) { 1526 struct bucket_array *buckets = gc_bucket_array(ca); 1527 struct bucket *g; 1528 1529 for_each_bucket(g, buckets) { 1530 if (metadata_only && 1531 (g->data_type == BCH_DATA_user || 1532 g->data_type == BCH_DATA_cached || 1533 g->data_type == BCH_DATA_parity)) 1534 continue; 1535 g->data_type = 0; 1536 g->dirty_sectors = 0; 1537 g->cached_sectors = 0; 1538 } 1539 } 1540 } 1541 1542 static int bch2_gc_write_reflink_key(struct btree_trans *trans, 1543 struct btree_iter *iter, 1544 struct bkey_s_c k, 1545 size_t *idx) 1546 { 1547 struct bch_fs *c = trans->c; 1548 const __le64 *refcount = bkey_refcount_c(k); 1549 struct printbuf buf = PRINTBUF; 1550 struct reflink_gc *r; 1551 int ret = 0; 1552 1553 if (!refcount) 1554 return 0; 1555 1556 while ((r = genradix_ptr(&c->reflink_gc_table, *idx)) && 1557 r->offset < k.k->p.offset) 1558 ++*idx; 1559 1560 if (!r || 1561 r->offset != k.k->p.offset || 1562 r->size != k.k->size) { 1563 bch_err(c, "unexpected inconsistency walking reflink table at gc finish"); 1564 return -EINVAL; 1565 } 1566 1567 if (fsck_err_on(r->refcount != le64_to_cpu(*refcount), c, 1568 reflink_v_refcount_wrong, 1569 "reflink key has wrong refcount:\n" 1570 " %s\n" 1571 " should be %u", 1572 (bch2_bkey_val_to_text(&buf, c, k), buf.buf), 1573 r->refcount)) { 1574 struct bkey_i *new = bch2_bkey_make_mut_noupdate(trans, k); 1575 ret = PTR_ERR_OR_ZERO(new); 1576 if (ret) 1577 return ret; 1578 1579 if (!r->refcount) 1580 new->k.type = KEY_TYPE_deleted; 1581 else 1582 *bkey_refcount(bkey_i_to_s(new)) = cpu_to_le64(r->refcount); 1583 ret = bch2_trans_update(trans, iter, new, 0); 1584 } 1585 fsck_err: 1586 printbuf_exit(&buf); 1587 return ret; 1588 } 1589 1590 static int bch2_gc_reflink_done(struct bch_fs *c, bool metadata_only) 1591 { 1592 size_t idx = 0; 1593 1594 if (metadata_only) 1595 return 0; 1596 1597 int ret = bch2_trans_run(c, 1598 for_each_btree_key_commit(trans, iter, 1599 BTREE_ID_reflink, POS_MIN, 1600 BTREE_ITER_PREFETCH, k, 1601 NULL, NULL, BCH_TRANS_COMMIT_no_enospc, 1602 bch2_gc_write_reflink_key(trans, &iter, k, &idx))); 1603 c->reflink_gc_nr = 0; 1604 return ret; 1605 } 1606 1607 static int bch2_gc_reflink_start(struct bch_fs *c, 1608 bool metadata_only) 1609 { 1610 1611 if (metadata_only) 1612 return 0; 1613 1614 c->reflink_gc_nr = 0; 1615 1616 int ret = bch2_trans_run(c, 1617 for_each_btree_key(trans, iter, BTREE_ID_reflink, POS_MIN, 1618 BTREE_ITER_PREFETCH, k, ({ 1619 const __le64 *refcount = bkey_refcount_c(k); 1620 1621 if (!refcount) 1622 continue; 1623 1624 struct reflink_gc *r = genradix_ptr_alloc(&c->reflink_gc_table, 1625 c->reflink_gc_nr++, GFP_KERNEL); 1626 if (!r) { 1627 ret = -BCH_ERR_ENOMEM_gc_reflink_start; 1628 break; 1629 } 1630 1631 r->offset = k.k->p.offset; 1632 r->size = k.k->size; 1633 r->refcount = 0; 1634 0; 1635 }))); 1636 1637 bch_err_fn(c, ret); 1638 return ret; 1639 } 1640 1641 static void bch2_gc_reflink_reset(struct bch_fs *c, bool metadata_only) 1642 { 1643 struct genradix_iter iter; 1644 struct reflink_gc *r; 1645 1646 genradix_for_each(&c->reflink_gc_table, iter, r) 1647 r->refcount = 0; 1648 } 1649 1650 static int bch2_gc_write_stripes_key(struct btree_trans *trans, 1651 struct btree_iter *iter, 1652 struct bkey_s_c k) 1653 { 1654 struct bch_fs *c = trans->c; 1655 struct printbuf buf = PRINTBUF; 1656 const struct bch_stripe *s; 1657 struct gc_stripe *m; 1658 bool bad = false; 1659 unsigned i; 1660 int ret = 0; 1661 1662 if (k.k->type != KEY_TYPE_stripe) 1663 return 0; 1664 1665 s = bkey_s_c_to_stripe(k).v; 1666 m = genradix_ptr(&c->gc_stripes, k.k->p.offset); 1667 1668 for (i = 0; i < s->nr_blocks; i++) { 1669 u32 old = stripe_blockcount_get(s, i); 1670 u32 new = (m ? m->block_sectors[i] : 0); 1671 1672 if (old != new) { 1673 prt_printf(&buf, "stripe block %u has wrong sector count: got %u, should be %u\n", 1674 i, old, new); 1675 bad = true; 1676 } 1677 } 1678 1679 if (bad) 1680 bch2_bkey_val_to_text(&buf, c, k); 1681 1682 if (fsck_err_on(bad, c, stripe_sector_count_wrong, 1683 "%s", buf.buf)) { 1684 struct bkey_i_stripe *new; 1685 1686 new = bch2_trans_kmalloc(trans, bkey_bytes(k.k)); 1687 ret = PTR_ERR_OR_ZERO(new); 1688 if (ret) 1689 return ret; 1690 1691 bkey_reassemble(&new->k_i, k); 1692 1693 for (i = 0; i < new->v.nr_blocks; i++) 1694 stripe_blockcount_set(&new->v, i, m ? m->block_sectors[i] : 0); 1695 1696 ret = bch2_trans_update(trans, iter, &new->k_i, 0); 1697 } 1698 fsck_err: 1699 printbuf_exit(&buf); 1700 return ret; 1701 } 1702 1703 static int bch2_gc_stripes_done(struct bch_fs *c, bool metadata_only) 1704 { 1705 if (metadata_only) 1706 return 0; 1707 1708 return bch2_trans_run(c, 1709 for_each_btree_key_commit(trans, iter, 1710 BTREE_ID_stripes, POS_MIN, 1711 BTREE_ITER_PREFETCH, k, 1712 NULL, NULL, BCH_TRANS_COMMIT_no_enospc, 1713 bch2_gc_write_stripes_key(trans, &iter, k))); 1714 } 1715 1716 static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only) 1717 { 1718 genradix_free(&c->gc_stripes); 1719 } 1720 1721 /** 1722 * bch2_gc - walk _all_ references to buckets, and recompute them: 1723 * 1724 * @c: filesystem object 1725 * @initial: are we in recovery? 1726 * @metadata_only: are we just checking metadata references, or everything? 1727 * 1728 * Returns: 0 on success, or standard errcode on failure 1729 * 1730 * Order matters here: 1731 * - Concurrent GC relies on the fact that we have a total ordering for 1732 * everything that GC walks - see gc_will_visit_node(), 1733 * gc_will_visit_root() 1734 * 1735 * - also, references move around in the course of index updates and 1736 * various other crap: everything needs to agree on the ordering 1737 * references are allowed to move around in - e.g., we're allowed to 1738 * start with a reference owned by an open_bucket (the allocator) and 1739 * move it to the btree, but not the reverse. 1740 * 1741 * This is necessary to ensure that gc doesn't miss references that 1742 * move around - if references move backwards in the ordering GC 1743 * uses, GC could skip past them 1744 */ 1745 int bch2_gc(struct bch_fs *c, bool initial, bool metadata_only) 1746 { 1747 unsigned iter = 0; 1748 int ret; 1749 1750 lockdep_assert_held(&c->state_lock); 1751 1752 down_write(&c->gc_lock); 1753 1754 bch2_btree_interior_updates_flush(c); 1755 1756 ret = bch2_gc_start(c) ?: 1757 bch2_gc_alloc_start(c, metadata_only) ?: 1758 bch2_gc_reflink_start(c, metadata_only); 1759 if (ret) 1760 goto out; 1761 again: 1762 gc_pos_set(c, gc_phase(GC_PHASE_START)); 1763 1764 bch2_mark_superblocks(c); 1765 1766 ret = bch2_gc_btrees(c, initial, metadata_only); 1767 1768 if (ret) 1769 goto out; 1770 1771 #if 0 1772 bch2_mark_pending_btree_node_frees(c); 1773 #endif 1774 c->gc_count++; 1775 1776 if (test_bit(BCH_FS_need_another_gc, &c->flags) || 1777 (!iter && bch2_test_restart_gc)) { 1778 if (iter++ > 2) { 1779 bch_info(c, "Unable to fix bucket gens, looping"); 1780 ret = -EINVAL; 1781 goto out; 1782 } 1783 1784 /* 1785 * XXX: make sure gens we fixed got saved 1786 */ 1787 bch_info(c, "Second GC pass needed, restarting:"); 1788 clear_bit(BCH_FS_need_another_gc, &c->flags); 1789 __gc_pos_set(c, gc_phase(GC_PHASE_NOT_RUNNING)); 1790 1791 bch2_gc_stripes_reset(c, metadata_only); 1792 bch2_gc_alloc_reset(c, metadata_only); 1793 bch2_gc_reflink_reset(c, metadata_only); 1794 ret = bch2_gc_reset(c); 1795 if (ret) 1796 goto out; 1797 1798 /* flush fsck errors, reset counters */ 1799 bch2_flush_fsck_errs(c); 1800 goto again; 1801 } 1802 out: 1803 if (!ret) { 1804 bch2_journal_block(&c->journal); 1805 1806 ret = bch2_gc_alloc_done(c, metadata_only) ?: 1807 bch2_gc_done(c, initial, metadata_only) ?: 1808 bch2_gc_stripes_done(c, metadata_only) ?: 1809 bch2_gc_reflink_done(c, metadata_only); 1810 1811 bch2_journal_unblock(&c->journal); 1812 } 1813 1814 percpu_down_write(&c->mark_lock); 1815 /* Indicates that gc is no longer in progress: */ 1816 __gc_pos_set(c, gc_phase(GC_PHASE_NOT_RUNNING)); 1817 1818 bch2_gc_free(c); 1819 percpu_up_write(&c->mark_lock); 1820 1821 up_write(&c->gc_lock); 1822 1823 /* 1824 * At startup, allocations can happen directly instead of via the 1825 * allocator thread - issue wakeup in case they blocked on gc_lock: 1826 */ 1827 closure_wake_up(&c->freelist_wait); 1828 bch_err_fn(c, ret); 1829 return ret; 1830 } 1831 1832 static int gc_btree_gens_key(struct btree_trans *trans, 1833 struct btree_iter *iter, 1834 struct bkey_s_c k) 1835 { 1836 struct bch_fs *c = trans->c; 1837 struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); 1838 struct bkey_i *u; 1839 int ret; 1840 1841 percpu_down_read(&c->mark_lock); 1842 bkey_for_each_ptr(ptrs, ptr) { 1843 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 1844 1845 if (ptr_stale(ca, ptr) > 16) { 1846 percpu_up_read(&c->mark_lock); 1847 goto update; 1848 } 1849 } 1850 1851 bkey_for_each_ptr(ptrs, ptr) { 1852 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 1853 u8 *gen = &ca->oldest_gen[PTR_BUCKET_NR(ca, ptr)]; 1854 1855 if (gen_after(*gen, ptr->gen)) 1856 *gen = ptr->gen; 1857 } 1858 percpu_up_read(&c->mark_lock); 1859 return 0; 1860 update: 1861 u = bch2_bkey_make_mut(trans, iter, &k, 0); 1862 ret = PTR_ERR_OR_ZERO(u); 1863 if (ret) 1864 return ret; 1865 1866 bch2_extent_normalize(c, bkey_i_to_s(u)); 1867 return 0; 1868 } 1869 1870 static int bch2_alloc_write_oldest_gen(struct btree_trans *trans, struct btree_iter *iter, 1871 struct bkey_s_c k) 1872 { 1873 struct bch_dev *ca = bch_dev_bkey_exists(trans->c, iter->pos.inode); 1874 struct bch_alloc_v4 a_convert; 1875 const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert); 1876 struct bkey_i_alloc_v4 *a_mut; 1877 int ret; 1878 1879 if (a->oldest_gen == ca->oldest_gen[iter->pos.offset]) 1880 return 0; 1881 1882 a_mut = bch2_alloc_to_v4_mut(trans, k); 1883 ret = PTR_ERR_OR_ZERO(a_mut); 1884 if (ret) 1885 return ret; 1886 1887 a_mut->v.oldest_gen = ca->oldest_gen[iter->pos.offset]; 1888 a_mut->v.data_type = alloc_data_type(a_mut->v, a_mut->v.data_type); 1889 1890 return bch2_trans_update(trans, iter, &a_mut->k_i, 0); 1891 } 1892 1893 int bch2_gc_gens(struct bch_fs *c) 1894 { 1895 u64 b, start_time = local_clock(); 1896 int ret; 1897 1898 /* 1899 * Ideally we would be using state_lock and not gc_lock here, but that 1900 * introduces a deadlock in the RO path - we currently take the state 1901 * lock at the start of going RO, thus the gc thread may get stuck: 1902 */ 1903 if (!mutex_trylock(&c->gc_gens_lock)) 1904 return 0; 1905 1906 trace_and_count(c, gc_gens_start, c); 1907 down_read(&c->gc_lock); 1908 1909 for_each_member_device(c, ca) { 1910 struct bucket_gens *gens = bucket_gens(ca); 1911 1912 BUG_ON(ca->oldest_gen); 1913 1914 ca->oldest_gen = kvmalloc(gens->nbuckets, GFP_KERNEL); 1915 if (!ca->oldest_gen) { 1916 percpu_ref_put(&ca->ref); 1917 ret = -BCH_ERR_ENOMEM_gc_gens; 1918 goto err; 1919 } 1920 1921 for (b = gens->first_bucket; 1922 b < gens->nbuckets; b++) 1923 ca->oldest_gen[b] = gens->b[b]; 1924 } 1925 1926 for (unsigned i = 0; i < BTREE_ID_NR; i++) 1927 if (btree_type_has_ptrs(i)) { 1928 c->gc_gens_btree = i; 1929 c->gc_gens_pos = POS_MIN; 1930 1931 ret = bch2_trans_run(c, 1932 for_each_btree_key_commit(trans, iter, i, 1933 POS_MIN, 1934 BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, 1935 k, 1936 NULL, NULL, 1937 BCH_TRANS_COMMIT_no_enospc, 1938 gc_btree_gens_key(trans, &iter, k))); 1939 if (ret) 1940 goto err; 1941 } 1942 1943 ret = bch2_trans_run(c, 1944 for_each_btree_key_commit(trans, iter, BTREE_ID_alloc, 1945 POS_MIN, 1946 BTREE_ITER_PREFETCH, 1947 k, 1948 NULL, NULL, 1949 BCH_TRANS_COMMIT_no_enospc, 1950 bch2_alloc_write_oldest_gen(trans, &iter, k))); 1951 if (ret) 1952 goto err; 1953 1954 c->gc_gens_btree = 0; 1955 c->gc_gens_pos = POS_MIN; 1956 1957 c->gc_count++; 1958 1959 bch2_time_stats_update(&c->times[BCH_TIME_btree_gc], start_time); 1960 trace_and_count(c, gc_gens_end, c); 1961 err: 1962 for_each_member_device(c, ca) { 1963 kvfree(ca->oldest_gen); 1964 ca->oldest_gen = NULL; 1965 } 1966 1967 up_read(&c->gc_lock); 1968 mutex_unlock(&c->gc_gens_lock); 1969 if (!bch2_err_matches(ret, EROFS)) 1970 bch_err_fn(c, ret); 1971 return ret; 1972 } 1973 1974 static int bch2_gc_thread(void *arg) 1975 { 1976 struct bch_fs *c = arg; 1977 struct io_clock *clock = &c->io_clock[WRITE]; 1978 unsigned long last = atomic64_read(&clock->now); 1979 unsigned last_kick = atomic_read(&c->kick_gc); 1980 1981 set_freezable(); 1982 1983 while (1) { 1984 while (1) { 1985 set_current_state(TASK_INTERRUPTIBLE); 1986 1987 if (kthread_should_stop()) { 1988 __set_current_state(TASK_RUNNING); 1989 return 0; 1990 } 1991 1992 if (atomic_read(&c->kick_gc) != last_kick) 1993 break; 1994 1995 if (c->btree_gc_periodic) { 1996 unsigned long next = last + c->capacity / 16; 1997 1998 if (atomic64_read(&clock->now) >= next) 1999 break; 2000 2001 bch2_io_clock_schedule_timeout(clock, next); 2002 } else { 2003 schedule(); 2004 } 2005 2006 try_to_freeze(); 2007 } 2008 __set_current_state(TASK_RUNNING); 2009 2010 last = atomic64_read(&clock->now); 2011 last_kick = atomic_read(&c->kick_gc); 2012 2013 /* 2014 * Full gc is currently incompatible with btree key cache: 2015 */ 2016 #if 0 2017 ret = bch2_gc(c, false, false); 2018 #else 2019 bch2_gc_gens(c); 2020 #endif 2021 debug_check_no_locks_held(); 2022 } 2023 2024 return 0; 2025 } 2026 2027 void bch2_gc_thread_stop(struct bch_fs *c) 2028 { 2029 struct task_struct *p; 2030 2031 p = c->gc_thread; 2032 c->gc_thread = NULL; 2033 2034 if (p) { 2035 kthread_stop(p); 2036 put_task_struct(p); 2037 } 2038 } 2039 2040 int bch2_gc_thread_start(struct bch_fs *c) 2041 { 2042 struct task_struct *p; 2043 2044 if (c->gc_thread) 2045 return 0; 2046 2047 p = kthread_create(bch2_gc_thread, c, "bch-gc/%s", c->name); 2048 if (IS_ERR(p)) { 2049 bch_err_fn(c, PTR_ERR(p)); 2050 return PTR_ERR(p); 2051 } 2052 2053 get_task_struct(p); 2054 c->gc_thread = p; 2055 wake_up_process(p); 2056 return 0; 2057 } 2058