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 "bkey_methods.h" 11 #include "bkey_buf.h" 12 #include "btree_journal_iter.h" 13 #include "btree_key_cache.h" 14 #include "btree_locking.h" 15 #include "btree_update_interior.h" 16 #include "btree_io.h" 17 #include "btree_gc.h" 18 #include "buckets.h" 19 #include "clock.h" 20 #include "debug.h" 21 #include "ec.h" 22 #include "error.h" 23 #include "extents.h" 24 #include "journal.h" 25 #include "keylist.h" 26 #include "move.h" 27 #include "recovery.h" 28 #include "reflink.h" 29 #include "replicas.h" 30 #include "super-io.h" 31 #include "trace.h" 32 33 #include <linux/slab.h> 34 #include <linux/bitops.h> 35 #include <linux/freezer.h> 36 #include <linux/kthread.h> 37 #include <linux/preempt.h> 38 #include <linux/rcupdate.h> 39 #include <linux/sched/task.h> 40 41 #define DROP_THIS_NODE 10 42 #define DROP_PREV_NODE 11 43 44 static struct bkey_s unsafe_bkey_s_c_to_s(struct bkey_s_c k) 45 { 46 return (struct bkey_s) {{{ 47 (struct bkey *) k.k, 48 (struct bch_val *) k.v 49 }}}; 50 } 51 52 static bool should_restart_for_topology_repair(struct bch_fs *c) 53 { 54 return c->opts.fix_errors != FSCK_FIX_no && 55 !(c->recovery_passes_complete & BIT_ULL(BCH_RECOVERY_PASS_check_topology)); 56 } 57 58 static inline void __gc_pos_set(struct bch_fs *c, struct gc_pos new_pos) 59 { 60 preempt_disable(); 61 write_seqcount_begin(&c->gc_pos_lock); 62 c->gc_pos = new_pos; 63 write_seqcount_end(&c->gc_pos_lock); 64 preempt_enable(); 65 } 66 67 static inline void gc_pos_set(struct bch_fs *c, struct gc_pos new_pos) 68 { 69 BUG_ON(gc_pos_cmp(new_pos, c->gc_pos) <= 0); 70 __gc_pos_set(c, new_pos); 71 } 72 73 /* 74 * Missing: if an interior btree node is empty, we need to do something - 75 * perhaps just kill it 76 */ 77 static int bch2_gc_check_topology(struct bch_fs *c, 78 struct btree *b, 79 struct bkey_buf *prev, 80 struct bkey_buf cur, 81 bool is_last) 82 { 83 struct bpos node_start = b->data->min_key; 84 struct bpos node_end = b->data->max_key; 85 struct bpos expected_start = bkey_deleted(&prev->k->k) 86 ? node_start 87 : bpos_successor(prev->k->k.p); 88 struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; 89 int ret = 0; 90 91 if (cur.k->k.type == KEY_TYPE_btree_ptr_v2) { 92 struct bkey_i_btree_ptr_v2 *bp = bkey_i_to_btree_ptr_v2(cur.k); 93 94 if (!bpos_eq(expected_start, bp->v.min_key)) { 95 bch2_topology_error(c); 96 97 if (bkey_deleted(&prev->k->k)) { 98 prt_printf(&buf1, "start of node: "); 99 bch2_bpos_to_text(&buf1, node_start); 100 } else { 101 bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(prev->k)); 102 } 103 bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(cur.k)); 104 105 if (__fsck_err(c, 106 FSCK_CAN_FIX| 107 FSCK_CAN_IGNORE| 108 FSCK_NO_RATELIMIT, 109 btree_node_topology_bad_min_key, 110 "btree node with incorrect min_key at btree %s level %u:\n" 111 " prev %s\n" 112 " cur %s", 113 bch2_btree_id_str(b->c.btree_id), b->c.level, 114 buf1.buf, buf2.buf) && should_restart_for_topology_repair(c)) { 115 bch_info(c, "Halting mark and sweep to start topology repair pass"); 116 ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); 117 goto err; 118 } else { 119 set_bit(BCH_FS_initial_gc_unfixed, &c->flags); 120 } 121 } 122 } 123 124 if (is_last && !bpos_eq(cur.k->k.p, node_end)) { 125 bch2_topology_error(c); 126 127 printbuf_reset(&buf1); 128 printbuf_reset(&buf2); 129 130 bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(cur.k)); 131 bch2_bpos_to_text(&buf2, node_end); 132 133 if (__fsck_err(c, FSCK_CAN_FIX|FSCK_CAN_IGNORE|FSCK_NO_RATELIMIT, 134 btree_node_topology_bad_max_key, 135 "btree node with incorrect max_key at btree %s level %u:\n" 136 " %s\n" 137 " expected %s", 138 bch2_btree_id_str(b->c.btree_id), b->c.level, 139 buf1.buf, buf2.buf) && 140 should_restart_for_topology_repair(c)) { 141 bch_info(c, "Halting mark and sweep to start topology repair pass"); 142 ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); 143 goto err; 144 } else { 145 set_bit(BCH_FS_initial_gc_unfixed, &c->flags); 146 } 147 } 148 149 bch2_bkey_buf_copy(prev, c, cur.k); 150 err: 151 fsck_err: 152 printbuf_exit(&buf2); 153 printbuf_exit(&buf1); 154 return ret; 155 } 156 157 static void btree_ptr_to_v2(struct btree *b, struct bkey_i_btree_ptr_v2 *dst) 158 { 159 switch (b->key.k.type) { 160 case KEY_TYPE_btree_ptr: { 161 struct bkey_i_btree_ptr *src = bkey_i_to_btree_ptr(&b->key); 162 163 dst->k.p = src->k.p; 164 dst->v.mem_ptr = 0; 165 dst->v.seq = b->data->keys.seq; 166 dst->v.sectors_written = 0; 167 dst->v.flags = 0; 168 dst->v.min_key = b->data->min_key; 169 set_bkey_val_bytes(&dst->k, sizeof(dst->v) + bkey_val_bytes(&src->k)); 170 memcpy(dst->v.start, src->v.start, bkey_val_bytes(&src->k)); 171 break; 172 } 173 case KEY_TYPE_btree_ptr_v2: 174 bkey_copy(&dst->k_i, &b->key); 175 break; 176 default: 177 BUG(); 178 } 179 } 180 181 static void bch2_btree_node_update_key_early(struct btree_trans *trans, 182 enum btree_id btree, unsigned level, 183 struct bkey_s_c old, struct bkey_i *new) 184 { 185 struct bch_fs *c = trans->c; 186 struct btree *b; 187 struct bkey_buf tmp; 188 int ret; 189 190 bch2_bkey_buf_init(&tmp); 191 bch2_bkey_buf_reassemble(&tmp, c, old); 192 193 b = bch2_btree_node_get_noiter(trans, tmp.k, btree, level, true); 194 if (!IS_ERR_OR_NULL(b)) { 195 mutex_lock(&c->btree_cache.lock); 196 197 bch2_btree_node_hash_remove(&c->btree_cache, b); 198 199 bkey_copy(&b->key, new); 200 ret = __bch2_btree_node_hash_insert(&c->btree_cache, b); 201 BUG_ON(ret); 202 203 mutex_unlock(&c->btree_cache.lock); 204 six_unlock_read(&b->c.lock); 205 } 206 207 bch2_bkey_buf_exit(&tmp, c); 208 } 209 210 static int set_node_min(struct bch_fs *c, struct btree *b, struct bpos new_min) 211 { 212 struct bkey_i_btree_ptr_v2 *new; 213 int ret; 214 215 new = kmalloc_array(BKEY_BTREE_PTR_U64s_MAX, sizeof(u64), GFP_KERNEL); 216 if (!new) 217 return -BCH_ERR_ENOMEM_gc_repair_key; 218 219 btree_ptr_to_v2(b, new); 220 b->data->min_key = new_min; 221 new->v.min_key = new_min; 222 SET_BTREE_PTR_RANGE_UPDATED(&new->v, true); 223 224 ret = bch2_journal_key_insert_take(c, b->c.btree_id, b->c.level + 1, &new->k_i); 225 if (ret) { 226 kfree(new); 227 return ret; 228 } 229 230 bch2_btree_node_drop_keys_outside_node(b); 231 bkey_copy(&b->key, &new->k_i); 232 return 0; 233 } 234 235 static int set_node_max(struct bch_fs *c, struct btree *b, struct bpos new_max) 236 { 237 struct bkey_i_btree_ptr_v2 *new; 238 int ret; 239 240 ret = bch2_journal_key_delete(c, b->c.btree_id, b->c.level + 1, b->key.k.p); 241 if (ret) 242 return ret; 243 244 new = kmalloc_array(BKEY_BTREE_PTR_U64s_MAX, sizeof(u64), GFP_KERNEL); 245 if (!new) 246 return -BCH_ERR_ENOMEM_gc_repair_key; 247 248 btree_ptr_to_v2(b, new); 249 b->data->max_key = new_max; 250 new->k.p = new_max; 251 SET_BTREE_PTR_RANGE_UPDATED(&new->v, true); 252 253 ret = bch2_journal_key_insert_take(c, b->c.btree_id, b->c.level + 1, &new->k_i); 254 if (ret) { 255 kfree(new); 256 return ret; 257 } 258 259 bch2_btree_node_drop_keys_outside_node(b); 260 261 mutex_lock(&c->btree_cache.lock); 262 bch2_btree_node_hash_remove(&c->btree_cache, b); 263 264 bkey_copy(&b->key, &new->k_i); 265 ret = __bch2_btree_node_hash_insert(&c->btree_cache, b); 266 BUG_ON(ret); 267 mutex_unlock(&c->btree_cache.lock); 268 return 0; 269 } 270 271 static int btree_repair_node_boundaries(struct bch_fs *c, struct btree *b, 272 struct btree *prev, struct btree *cur) 273 { 274 struct bpos expected_start = !prev 275 ? b->data->min_key 276 : bpos_successor(prev->key.k.p); 277 struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; 278 int ret = 0; 279 280 if (!prev) { 281 prt_printf(&buf1, "start of node: "); 282 bch2_bpos_to_text(&buf1, b->data->min_key); 283 } else { 284 bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(&prev->key)); 285 } 286 287 bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(&cur->key)); 288 289 if (prev && 290 bpos_gt(expected_start, cur->data->min_key) && 291 BTREE_NODE_SEQ(cur->data) > BTREE_NODE_SEQ(prev->data)) { 292 /* cur overwrites prev: */ 293 294 if (mustfix_fsck_err_on(bpos_ge(prev->data->min_key, 295 cur->data->min_key), c, 296 btree_node_topology_overwritten_by_next_node, 297 "btree node overwritten by next node at btree %s level %u:\n" 298 " node %s\n" 299 " next %s", 300 bch2_btree_id_str(b->c.btree_id), b->c.level, 301 buf1.buf, buf2.buf)) { 302 ret = DROP_PREV_NODE; 303 goto out; 304 } 305 306 if (mustfix_fsck_err_on(!bpos_eq(prev->key.k.p, 307 bpos_predecessor(cur->data->min_key)), c, 308 btree_node_topology_bad_max_key, 309 "btree node with incorrect max_key at btree %s level %u:\n" 310 " node %s\n" 311 " next %s", 312 bch2_btree_id_str(b->c.btree_id), b->c.level, 313 buf1.buf, buf2.buf)) 314 ret = set_node_max(c, prev, 315 bpos_predecessor(cur->data->min_key)); 316 } else { 317 /* prev overwrites cur: */ 318 319 if (mustfix_fsck_err_on(bpos_ge(expected_start, 320 cur->data->max_key), c, 321 btree_node_topology_overwritten_by_prev_node, 322 "btree node overwritten by prev node at btree %s level %u:\n" 323 " prev %s\n" 324 " node %s", 325 bch2_btree_id_str(b->c.btree_id), b->c.level, 326 buf1.buf, buf2.buf)) { 327 ret = DROP_THIS_NODE; 328 goto out; 329 } 330 331 if (mustfix_fsck_err_on(!bpos_eq(expected_start, cur->data->min_key), c, 332 btree_node_topology_bad_min_key, 333 "btree node with incorrect min_key at btree %s level %u:\n" 334 " prev %s\n" 335 " node %s", 336 bch2_btree_id_str(b->c.btree_id), b->c.level, 337 buf1.buf, buf2.buf)) 338 ret = set_node_min(c, cur, expected_start); 339 } 340 out: 341 fsck_err: 342 printbuf_exit(&buf2); 343 printbuf_exit(&buf1); 344 return ret; 345 } 346 347 static int btree_repair_node_end(struct bch_fs *c, struct btree *b, 348 struct btree *child) 349 { 350 struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF; 351 int ret = 0; 352 353 bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(&child->key)); 354 bch2_bpos_to_text(&buf2, b->key.k.p); 355 356 if (mustfix_fsck_err_on(!bpos_eq(child->key.k.p, b->key.k.p), c, 357 btree_node_topology_bad_max_key, 358 "btree node with incorrect max_key at btree %s level %u:\n" 359 " %s\n" 360 " expected %s", 361 bch2_btree_id_str(b->c.btree_id), b->c.level, 362 buf1.buf, buf2.buf)) { 363 ret = set_node_max(c, child, b->key.k.p); 364 if (ret) 365 goto err; 366 } 367 err: 368 fsck_err: 369 printbuf_exit(&buf2); 370 printbuf_exit(&buf1); 371 return ret; 372 } 373 374 static int bch2_btree_repair_topology_recurse(struct btree_trans *trans, struct btree *b) 375 { 376 struct bch_fs *c = trans->c; 377 struct btree_and_journal_iter iter; 378 struct bkey_s_c k; 379 struct bkey_buf prev_k, cur_k; 380 struct btree *prev = NULL, *cur = NULL; 381 bool have_child, dropped_children = false; 382 struct printbuf buf = PRINTBUF; 383 int ret = 0; 384 385 if (!b->c.level) 386 return 0; 387 again: 388 prev = NULL; 389 have_child = dropped_children = false; 390 bch2_bkey_buf_init(&prev_k); 391 bch2_bkey_buf_init(&cur_k); 392 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 393 iter.prefetch = true; 394 395 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 396 BUG_ON(bpos_lt(k.k->p, b->data->min_key)); 397 BUG_ON(bpos_gt(k.k->p, b->data->max_key)); 398 399 bch2_btree_and_journal_iter_advance(&iter); 400 bch2_bkey_buf_reassemble(&cur_k, c, k); 401 402 cur = bch2_btree_node_get_noiter(trans, cur_k.k, 403 b->c.btree_id, b->c.level - 1, 404 false); 405 ret = PTR_ERR_OR_ZERO(cur); 406 407 printbuf_reset(&buf); 408 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur_k.k)); 409 410 if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO), c, 411 btree_node_unreadable, 412 "Topology repair: unreadable btree node at btree %s level %u:\n" 413 " %s", 414 bch2_btree_id_str(b->c.btree_id), 415 b->c.level - 1, 416 buf.buf)) { 417 bch2_btree_node_evict(trans, cur_k.k); 418 ret = bch2_journal_key_delete(c, b->c.btree_id, 419 b->c.level, cur_k.k->k.p); 420 cur = NULL; 421 if (ret) 422 break; 423 continue; 424 } 425 426 bch_err_msg(c, ret, "getting btree node"); 427 if (ret) 428 break; 429 430 ret = btree_repair_node_boundaries(c, b, prev, cur); 431 432 if (ret == DROP_THIS_NODE) { 433 six_unlock_read(&cur->c.lock); 434 bch2_btree_node_evict(trans, cur_k.k); 435 ret = bch2_journal_key_delete(c, b->c.btree_id, 436 b->c.level, cur_k.k->k.p); 437 cur = NULL; 438 if (ret) 439 break; 440 continue; 441 } 442 443 if (prev) 444 six_unlock_read(&prev->c.lock); 445 prev = NULL; 446 447 if (ret == DROP_PREV_NODE) { 448 bch2_btree_node_evict(trans, prev_k.k); 449 ret = bch2_journal_key_delete(c, b->c.btree_id, 450 b->c.level, prev_k.k->k.p); 451 if (ret) 452 break; 453 454 bch2_btree_and_journal_iter_exit(&iter); 455 bch2_bkey_buf_exit(&prev_k, c); 456 bch2_bkey_buf_exit(&cur_k, c); 457 goto again; 458 } else if (ret) 459 break; 460 461 prev = cur; 462 cur = NULL; 463 bch2_bkey_buf_copy(&prev_k, c, cur_k.k); 464 } 465 466 if (!ret && !IS_ERR_OR_NULL(prev)) { 467 BUG_ON(cur); 468 ret = btree_repair_node_end(c, b, prev); 469 } 470 471 if (!IS_ERR_OR_NULL(prev)) 472 six_unlock_read(&prev->c.lock); 473 prev = NULL; 474 if (!IS_ERR_OR_NULL(cur)) 475 six_unlock_read(&cur->c.lock); 476 cur = NULL; 477 478 if (ret) 479 goto err; 480 481 bch2_btree_and_journal_iter_exit(&iter); 482 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 483 iter.prefetch = true; 484 485 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 486 bch2_bkey_buf_reassemble(&cur_k, c, k); 487 bch2_btree_and_journal_iter_advance(&iter); 488 489 cur = bch2_btree_node_get_noiter(trans, cur_k.k, 490 b->c.btree_id, b->c.level - 1, 491 false); 492 ret = PTR_ERR_OR_ZERO(cur); 493 494 bch_err_msg(c, ret, "getting btree node"); 495 if (ret) 496 goto err; 497 498 ret = bch2_btree_repair_topology_recurse(trans, cur); 499 six_unlock_read(&cur->c.lock); 500 cur = NULL; 501 502 if (ret == DROP_THIS_NODE) { 503 bch2_btree_node_evict(trans, cur_k.k); 504 ret = bch2_journal_key_delete(c, b->c.btree_id, 505 b->c.level, cur_k.k->k.p); 506 dropped_children = true; 507 } 508 509 if (ret) 510 goto err; 511 512 have_child = true; 513 } 514 515 printbuf_reset(&buf); 516 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(&b->key)); 517 518 if (mustfix_fsck_err_on(!have_child, c, 519 btree_node_topology_interior_node_empty, 520 "empty interior btree node at btree %s level %u\n" 521 " %s", 522 bch2_btree_id_str(b->c.btree_id), 523 b->c.level, buf.buf)) 524 ret = DROP_THIS_NODE; 525 err: 526 fsck_err: 527 if (!IS_ERR_OR_NULL(prev)) 528 six_unlock_read(&prev->c.lock); 529 if (!IS_ERR_OR_NULL(cur)) 530 six_unlock_read(&cur->c.lock); 531 532 bch2_btree_and_journal_iter_exit(&iter); 533 bch2_bkey_buf_exit(&prev_k, c); 534 bch2_bkey_buf_exit(&cur_k, c); 535 536 if (!ret && dropped_children) 537 goto again; 538 539 printbuf_exit(&buf); 540 return ret; 541 } 542 543 int bch2_check_topology(struct bch_fs *c) 544 { 545 struct btree_trans *trans = bch2_trans_get(c); 546 struct btree *b; 547 unsigned i; 548 int ret = 0; 549 550 for (i = 0; i < btree_id_nr_alive(c) && !ret; i++) { 551 struct btree_root *r = bch2_btree_id_root(c, i); 552 553 if (!r->alive) 554 continue; 555 556 b = r->b; 557 if (btree_node_fake(b)) 558 continue; 559 560 btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_read); 561 ret = bch2_btree_repair_topology_recurse(trans, b); 562 six_unlock_read(&b->c.lock); 563 564 if (ret == DROP_THIS_NODE) { 565 bch_err(c, "empty btree root - repair unimplemented"); 566 ret = -BCH_ERR_fsck_repair_unimplemented; 567 } 568 } 569 570 bch2_trans_put(trans); 571 572 return ret; 573 } 574 575 static int bch2_check_fix_ptrs(struct btree_trans *trans, enum btree_id btree_id, 576 unsigned level, bool is_root, 577 struct bkey_s_c *k) 578 { 579 struct bch_fs *c = trans->c; 580 struct bkey_ptrs_c ptrs_c = bch2_bkey_ptrs_c(*k); 581 const union bch_extent_entry *entry_c; 582 struct extent_ptr_decoded p = { 0 }; 583 bool do_update = false; 584 struct printbuf buf = PRINTBUF; 585 int ret = 0; 586 587 /* 588 * XXX 589 * use check_bucket_ref here 590 */ 591 bkey_for_each_ptr_decode(k->k, ptrs_c, p, entry_c) { 592 struct bch_dev *ca = bch_dev_bkey_exists(c, p.ptr.dev); 593 struct bucket *g = PTR_GC_BUCKET(ca, &p.ptr); 594 enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, &entry_c->ptr); 595 596 if (fsck_err_on(!g->gen_valid, 597 c, ptr_to_missing_alloc_key, 598 "bucket %u:%zu data type %s ptr gen %u missing in alloc btree\n" 599 "while marking %s", 600 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 601 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 602 p.ptr.gen, 603 (printbuf_reset(&buf), 604 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) { 605 if (!p.ptr.cached) { 606 g->gen_valid = true; 607 g->gen = p.ptr.gen; 608 } else { 609 do_update = true; 610 } 611 } 612 613 if (fsck_err_on(gen_cmp(p.ptr.gen, g->gen) > 0, 614 c, ptr_gen_newer_than_bucket_gen, 615 "bucket %u:%zu data type %s ptr gen in the future: %u > %u\n" 616 "while marking %s", 617 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 618 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 619 p.ptr.gen, g->gen, 620 (printbuf_reset(&buf), 621 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) { 622 if (!p.ptr.cached) { 623 g->gen_valid = true; 624 g->gen = p.ptr.gen; 625 g->data_type = 0; 626 g->dirty_sectors = 0; 627 g->cached_sectors = 0; 628 set_bit(BCH_FS_need_another_gc, &c->flags); 629 } else { 630 do_update = true; 631 } 632 } 633 634 if (fsck_err_on(gen_cmp(g->gen, p.ptr.gen) > BUCKET_GC_GEN_MAX, 635 c, ptr_gen_newer_than_bucket_gen, 636 "bucket %u:%zu gen %u data type %s: ptr gen %u too stale\n" 637 "while marking %s", 638 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), g->gen, 639 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 640 p.ptr.gen, 641 (printbuf_reset(&buf), 642 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 643 do_update = true; 644 645 if (fsck_err_on(!p.ptr.cached && gen_cmp(p.ptr.gen, g->gen) < 0, 646 c, stale_dirty_ptr, 647 "bucket %u:%zu data type %s stale dirty ptr: %u < %u\n" 648 "while marking %s", 649 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 650 bch2_data_type_str(ptr_data_type(k->k, &p.ptr)), 651 p.ptr.gen, g->gen, 652 (printbuf_reset(&buf), 653 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 654 do_update = true; 655 656 if (data_type != BCH_DATA_btree && p.ptr.gen != g->gen) 657 continue; 658 659 if (fsck_err_on(bucket_data_type(g->data_type) && 660 bucket_data_type(g->data_type) != data_type, c, 661 ptr_bucket_data_type_mismatch, 662 "bucket %u:%zu different types of data in same bucket: %s, %s\n" 663 "while marking %s", 664 p.ptr.dev, PTR_BUCKET_NR(ca, &p.ptr), 665 bch2_data_type_str(g->data_type), 666 bch2_data_type_str(data_type), 667 (printbuf_reset(&buf), 668 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) { 669 if (data_type == BCH_DATA_btree) { 670 g->data_type = data_type; 671 set_bit(BCH_FS_need_another_gc, &c->flags); 672 } else { 673 do_update = true; 674 } 675 } 676 677 if (p.has_ec) { 678 struct gc_stripe *m = genradix_ptr(&c->gc_stripes, p.ec.idx); 679 680 if (fsck_err_on(!m || !m->alive, c, 681 ptr_to_missing_stripe, 682 "pointer to nonexistent stripe %llu\n" 683 "while marking %s", 684 (u64) p.ec.idx, 685 (printbuf_reset(&buf), 686 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 687 do_update = true; 688 689 if (fsck_err_on(m && m->alive && !bch2_ptr_matches_stripe_m(m, p), c, 690 ptr_to_incorrect_stripe, 691 "pointer does not match stripe %llu\n" 692 "while marking %s", 693 (u64) p.ec.idx, 694 (printbuf_reset(&buf), 695 bch2_bkey_val_to_text(&buf, c, *k), buf.buf))) 696 do_update = true; 697 } 698 } 699 700 if (do_update) { 701 struct bkey_ptrs ptrs; 702 union bch_extent_entry *entry; 703 struct bch_extent_ptr *ptr; 704 struct bkey_i *new; 705 706 if (is_root) { 707 bch_err(c, "cannot update btree roots yet"); 708 ret = -EINVAL; 709 goto err; 710 } 711 712 new = kmalloc(bkey_bytes(k->k), GFP_KERNEL); 713 if (!new) { 714 ret = -BCH_ERR_ENOMEM_gc_repair_key; 715 bch_err_msg(c, ret, "allocating new key"); 716 goto err; 717 } 718 719 bkey_reassemble(new, *k); 720 721 if (level) { 722 /* 723 * We don't want to drop btree node pointers - if the 724 * btree node isn't there anymore, the read path will 725 * sort it out: 726 */ 727 ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); 728 bkey_for_each_ptr(ptrs, ptr) { 729 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 730 struct bucket *g = PTR_GC_BUCKET(ca, ptr); 731 732 ptr->gen = g->gen; 733 } 734 } else { 735 bch2_bkey_drop_ptrs(bkey_i_to_s(new), ptr, ({ 736 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 737 struct bucket *g = PTR_GC_BUCKET(ca, ptr); 738 enum bch_data_type data_type = bch2_bkey_ptr_data_type(*k, ptr); 739 740 (ptr->cached && 741 (!g->gen_valid || gen_cmp(ptr->gen, g->gen) > 0)) || 742 (!ptr->cached && 743 gen_cmp(ptr->gen, g->gen) < 0) || 744 gen_cmp(g->gen, ptr->gen) > BUCKET_GC_GEN_MAX || 745 (g->data_type && 746 g->data_type != data_type); 747 })); 748 again: 749 ptrs = bch2_bkey_ptrs(bkey_i_to_s(new)); 750 bkey_extent_entry_for_each(ptrs, entry) { 751 if (extent_entry_type(entry) == BCH_EXTENT_ENTRY_stripe_ptr) { 752 struct gc_stripe *m = genradix_ptr(&c->gc_stripes, 753 entry->stripe_ptr.idx); 754 union bch_extent_entry *next_ptr; 755 756 bkey_extent_entry_for_each_from(ptrs, next_ptr, entry) 757 if (extent_entry_type(next_ptr) == BCH_EXTENT_ENTRY_ptr) 758 goto found; 759 next_ptr = NULL; 760 found: 761 if (!next_ptr) { 762 bch_err(c, "aieee, found stripe ptr with no data ptr"); 763 continue; 764 } 765 766 if (!m || !m->alive || 767 !__bch2_ptr_matches_stripe(&m->ptrs[entry->stripe_ptr.block], 768 &next_ptr->ptr, 769 m->sectors)) { 770 bch2_bkey_extent_entry_drop(new, entry); 771 goto again; 772 } 773 } 774 } 775 } 776 777 ret = bch2_journal_key_insert_take(c, btree_id, level, new); 778 if (ret) { 779 kfree(new); 780 goto err; 781 } 782 783 if (level) 784 bch2_btree_node_update_key_early(trans, btree_id, level - 1, *k, new); 785 786 if (0) { 787 printbuf_reset(&buf); 788 bch2_bkey_val_to_text(&buf, c, *k); 789 bch_info(c, "updated %s", buf.buf); 790 791 printbuf_reset(&buf); 792 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(new)); 793 bch_info(c, "new key %s", buf.buf); 794 } 795 796 *k = bkey_i_to_s_c(new); 797 } 798 err: 799 fsck_err: 800 printbuf_exit(&buf); 801 return ret; 802 } 803 804 /* marking of btree keys/nodes: */ 805 806 static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id, 807 unsigned level, bool is_root, 808 struct bkey_s_c *k, 809 bool initial) 810 { 811 struct bch_fs *c = trans->c; 812 struct bkey deleted = KEY(0, 0, 0); 813 struct bkey_s_c old = (struct bkey_s_c) { &deleted, NULL }; 814 int ret = 0; 815 816 deleted.p = k->k->p; 817 818 if (initial) { 819 BUG_ON(bch2_journal_seq_verify && 820 k->k->version.lo > atomic64_read(&c->journal.seq)); 821 822 ret = bch2_check_fix_ptrs(trans, btree_id, level, is_root, k); 823 if (ret) 824 goto err; 825 826 if (fsck_err_on(k->k->version.lo > atomic64_read(&c->key_version), c, 827 bkey_version_in_future, 828 "key version number higher than recorded: %llu > %llu", 829 k->k->version.lo, 830 atomic64_read(&c->key_version))) 831 atomic64_set(&c->key_version, k->k->version.lo); 832 } 833 834 ret = commit_do(trans, NULL, NULL, 0, 835 bch2_key_trigger(trans, btree_id, level, old, unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC)); 836 fsck_err: 837 err: 838 bch_err_fn(c, ret); 839 return ret; 840 } 841 842 static int btree_gc_mark_node(struct btree_trans *trans, struct btree *b, bool initial) 843 { 844 struct bch_fs *c = trans->c; 845 struct btree_node_iter iter; 846 struct bkey unpacked; 847 struct bkey_s_c k; 848 struct bkey_buf prev, cur; 849 int ret = 0; 850 851 if (!btree_node_type_needs_gc(btree_node_type(b))) 852 return 0; 853 854 bch2_btree_node_iter_init_from_start(&iter, b); 855 bch2_bkey_buf_init(&prev); 856 bch2_bkey_buf_init(&cur); 857 bkey_init(&prev.k->k); 858 859 while ((k = bch2_btree_node_iter_peek_unpack(&iter, b, &unpacked)).k) { 860 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, false, 861 &k, initial); 862 if (ret) 863 break; 864 865 bch2_btree_node_iter_advance(&iter, b); 866 867 if (b->c.level) { 868 bch2_bkey_buf_reassemble(&cur, c, k); 869 870 ret = bch2_gc_check_topology(c, b, &prev, cur, 871 bch2_btree_node_iter_end(&iter)); 872 if (ret) 873 break; 874 } 875 } 876 877 bch2_bkey_buf_exit(&cur, c); 878 bch2_bkey_buf_exit(&prev, c); 879 return ret; 880 } 881 882 static int bch2_gc_btree(struct btree_trans *trans, enum btree_id btree_id, 883 bool initial, bool metadata_only) 884 { 885 struct bch_fs *c = trans->c; 886 struct btree_iter iter; 887 struct btree *b; 888 unsigned depth = metadata_only ? 1 : 0; 889 int ret = 0; 890 891 gc_pos_set(c, gc_pos_btree(btree_id, POS_MIN, 0)); 892 893 __for_each_btree_node(trans, iter, btree_id, POS_MIN, 894 0, depth, BTREE_ITER_PREFETCH, b, ret) { 895 bch2_verify_btree_nr_keys(b); 896 897 gc_pos_set(c, gc_pos_btree_node(b)); 898 899 ret = btree_gc_mark_node(trans, b, initial); 900 if (ret) 901 break; 902 } 903 bch2_trans_iter_exit(trans, &iter); 904 905 if (ret) 906 return ret; 907 908 mutex_lock(&c->btree_root_lock); 909 b = bch2_btree_id_root(c, btree_id)->b; 910 if (!btree_node_fake(b)) { 911 struct bkey_s_c k = bkey_i_to_s_c(&b->key); 912 913 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1, 914 true, &k, initial); 915 } 916 gc_pos_set(c, gc_pos_btree_root(b->c.btree_id)); 917 mutex_unlock(&c->btree_root_lock); 918 919 return ret; 920 } 921 922 static int bch2_gc_btree_init_recurse(struct btree_trans *trans, struct btree *b, 923 unsigned target_depth) 924 { 925 struct bch_fs *c = trans->c; 926 struct btree_and_journal_iter iter; 927 struct bkey_s_c k; 928 struct bkey_buf cur, prev; 929 struct printbuf buf = PRINTBUF; 930 int ret = 0; 931 932 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 933 bch2_bkey_buf_init(&prev); 934 bch2_bkey_buf_init(&cur); 935 bkey_init(&prev.k->k); 936 937 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 938 BUG_ON(bpos_lt(k.k->p, b->data->min_key)); 939 BUG_ON(bpos_gt(k.k->p, b->data->max_key)); 940 941 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level, 942 false, &k, true); 943 if (ret) 944 goto fsck_err; 945 946 if (b->c.level) { 947 bch2_bkey_buf_reassemble(&cur, c, k); 948 k = bkey_i_to_s_c(cur.k); 949 950 bch2_btree_and_journal_iter_advance(&iter); 951 952 ret = bch2_gc_check_topology(c, b, 953 &prev, cur, 954 !bch2_btree_and_journal_iter_peek(&iter).k); 955 if (ret) 956 goto fsck_err; 957 } else { 958 bch2_btree_and_journal_iter_advance(&iter); 959 } 960 } 961 962 if (b->c.level > target_depth) { 963 bch2_btree_and_journal_iter_exit(&iter); 964 bch2_btree_and_journal_iter_init_node_iter(trans, &iter, b); 965 iter.prefetch = true; 966 967 while ((k = bch2_btree_and_journal_iter_peek(&iter)).k) { 968 struct btree *child; 969 970 bch2_bkey_buf_reassemble(&cur, c, k); 971 bch2_btree_and_journal_iter_advance(&iter); 972 973 child = bch2_btree_node_get_noiter(trans, cur.k, 974 b->c.btree_id, b->c.level - 1, 975 false); 976 ret = PTR_ERR_OR_ZERO(child); 977 978 if (bch2_err_matches(ret, EIO)) { 979 bch2_topology_error(c); 980 981 if (__fsck_err(c, 982 FSCK_CAN_FIX| 983 FSCK_CAN_IGNORE| 984 FSCK_NO_RATELIMIT, 985 btree_node_read_error, 986 "Unreadable btree node at btree %s level %u:\n" 987 " %s", 988 bch2_btree_id_str(b->c.btree_id), 989 b->c.level - 1, 990 (printbuf_reset(&buf), 991 bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur.k)), buf.buf)) && 992 should_restart_for_topology_repair(c)) { 993 bch_info(c, "Halting mark and sweep to start topology repair pass"); 994 ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_check_topology); 995 goto fsck_err; 996 } else { 997 /* Continue marking when opted to not 998 * fix the error: */ 999 ret = 0; 1000 set_bit(BCH_FS_initial_gc_unfixed, &c->flags); 1001 continue; 1002 } 1003 } else if (ret) { 1004 bch_err_msg(c, ret, "getting btree node"); 1005 break; 1006 } 1007 1008 ret = bch2_gc_btree_init_recurse(trans, child, 1009 target_depth); 1010 six_unlock_read(&child->c.lock); 1011 1012 if (ret) 1013 break; 1014 } 1015 } 1016 fsck_err: 1017 bch2_bkey_buf_exit(&cur, c); 1018 bch2_bkey_buf_exit(&prev, c); 1019 bch2_btree_and_journal_iter_exit(&iter); 1020 printbuf_exit(&buf); 1021 return ret; 1022 } 1023 1024 static int bch2_gc_btree_init(struct btree_trans *trans, 1025 enum btree_id btree_id, 1026 bool metadata_only) 1027 { 1028 struct bch_fs *c = trans->c; 1029 struct btree *b; 1030 unsigned target_depth = metadata_only ? 1 : 0; 1031 struct printbuf buf = PRINTBUF; 1032 int ret = 0; 1033 1034 b = bch2_btree_id_root(c, btree_id)->b; 1035 1036 if (btree_node_fake(b)) 1037 return 0; 1038 1039 six_lock_read(&b->c.lock, NULL, NULL); 1040 printbuf_reset(&buf); 1041 bch2_bpos_to_text(&buf, b->data->min_key); 1042 if (mustfix_fsck_err_on(!bpos_eq(b->data->min_key, POS_MIN), c, 1043 btree_root_bad_min_key, 1044 "btree root with incorrect min_key: %s", buf.buf)) { 1045 bch_err(c, "repair unimplemented"); 1046 ret = -BCH_ERR_fsck_repair_unimplemented; 1047 goto fsck_err; 1048 } 1049 1050 printbuf_reset(&buf); 1051 bch2_bpos_to_text(&buf, b->data->max_key); 1052 if (mustfix_fsck_err_on(!bpos_eq(b->data->max_key, SPOS_MAX), c, 1053 btree_root_bad_max_key, 1054 "btree root with incorrect max_key: %s", buf.buf)) { 1055 bch_err(c, "repair unimplemented"); 1056 ret = -BCH_ERR_fsck_repair_unimplemented; 1057 goto fsck_err; 1058 } 1059 1060 if (b->c.level >= target_depth) 1061 ret = bch2_gc_btree_init_recurse(trans, b, target_depth); 1062 1063 if (!ret) { 1064 struct bkey_s_c k = bkey_i_to_s_c(&b->key); 1065 1066 ret = bch2_gc_mark_key(trans, b->c.btree_id, b->c.level + 1, true, 1067 &k, true); 1068 } 1069 fsck_err: 1070 six_unlock_read(&b->c.lock); 1071 1072 bch_err_fn(c, ret); 1073 printbuf_exit(&buf); 1074 return ret; 1075 } 1076 1077 static inline int btree_id_gc_phase_cmp(enum btree_id l, enum btree_id r) 1078 { 1079 return (int) btree_id_to_gc_phase(l) - 1080 (int) btree_id_to_gc_phase(r); 1081 } 1082 1083 static int bch2_gc_btrees(struct bch_fs *c, bool initial, bool metadata_only) 1084 { 1085 struct btree_trans *trans = bch2_trans_get(c); 1086 enum btree_id ids[BTREE_ID_NR]; 1087 unsigned i; 1088 int ret = 0; 1089 1090 for (i = 0; i < BTREE_ID_NR; i++) 1091 ids[i] = i; 1092 bubble_sort(ids, BTREE_ID_NR, btree_id_gc_phase_cmp); 1093 1094 for (i = 0; i < BTREE_ID_NR && !ret; i++) 1095 ret = initial 1096 ? bch2_gc_btree_init(trans, ids[i], metadata_only) 1097 : bch2_gc_btree(trans, ids[i], initial, metadata_only); 1098 1099 for (i = BTREE_ID_NR; i < btree_id_nr_alive(c) && !ret; i++) { 1100 if (!bch2_btree_id_root(c, i)->alive) 1101 continue; 1102 1103 ret = initial 1104 ? bch2_gc_btree_init(trans, i, metadata_only) 1105 : bch2_gc_btree(trans, i, initial, metadata_only); 1106 } 1107 1108 bch2_trans_put(trans); 1109 bch_err_fn(c, ret); 1110 return ret; 1111 } 1112 1113 static void mark_metadata_sectors(struct bch_fs *c, struct bch_dev *ca, 1114 u64 start, u64 end, 1115 enum bch_data_type type, 1116 unsigned flags) 1117 { 1118 u64 b = sector_to_bucket(ca, start); 1119 1120 do { 1121 unsigned sectors = 1122 min_t(u64, bucket_to_sector(ca, b + 1), end) - start; 1123 1124 bch2_mark_metadata_bucket(c, ca, b, type, sectors, 1125 gc_phase(GC_PHASE_SB), flags); 1126 b++; 1127 start += sectors; 1128 } while (start < end); 1129 } 1130 1131 static void bch2_mark_dev_superblock(struct bch_fs *c, struct bch_dev *ca, 1132 unsigned flags) 1133 { 1134 struct bch_sb_layout *layout = &ca->disk_sb.sb->layout; 1135 unsigned i; 1136 u64 b; 1137 1138 for (i = 0; i < layout->nr_superblocks; i++) { 1139 u64 offset = le64_to_cpu(layout->sb_offset[i]); 1140 1141 if (offset == BCH_SB_SECTOR) 1142 mark_metadata_sectors(c, ca, 0, BCH_SB_SECTOR, 1143 BCH_DATA_sb, flags); 1144 1145 mark_metadata_sectors(c, ca, offset, 1146 offset + (1 << layout->sb_max_size_bits), 1147 BCH_DATA_sb, flags); 1148 } 1149 1150 for (i = 0; i < ca->journal.nr; i++) { 1151 b = ca->journal.buckets[i]; 1152 bch2_mark_metadata_bucket(c, ca, b, BCH_DATA_journal, 1153 ca->mi.bucket_size, 1154 gc_phase(GC_PHASE_SB), flags); 1155 } 1156 } 1157 1158 static void bch2_mark_superblocks(struct bch_fs *c) 1159 { 1160 mutex_lock(&c->sb_lock); 1161 gc_pos_set(c, gc_phase(GC_PHASE_SB)); 1162 1163 for_each_online_member(c, ca) 1164 bch2_mark_dev_superblock(c, ca, BTREE_TRIGGER_GC); 1165 mutex_unlock(&c->sb_lock); 1166 } 1167 1168 #if 0 1169 /* Also see bch2_pending_btree_node_free_insert_done() */ 1170 static void bch2_mark_pending_btree_node_frees(struct bch_fs *c) 1171 { 1172 struct btree_update *as; 1173 struct pending_btree_node_free *d; 1174 1175 mutex_lock(&c->btree_interior_update_lock); 1176 gc_pos_set(c, gc_phase(GC_PHASE_PENDING_DELETE)); 1177 1178 for_each_pending_btree_node_free(c, as, d) 1179 if (d->index_update_done) 1180 bch2_mark_key(c, bkey_i_to_s_c(&d->key), BTREE_TRIGGER_GC); 1181 1182 mutex_unlock(&c->btree_interior_update_lock); 1183 } 1184 #endif 1185 1186 static void bch2_gc_free(struct bch_fs *c) 1187 { 1188 genradix_free(&c->reflink_gc_table); 1189 genradix_free(&c->gc_stripes); 1190 1191 for_each_member_device(c, ca) { 1192 kvfree(rcu_dereference_protected(ca->buckets_gc, 1)); 1193 ca->buckets_gc = NULL; 1194 1195 free_percpu(ca->usage_gc); 1196 ca->usage_gc = NULL; 1197 } 1198 1199 free_percpu(c->usage_gc); 1200 c->usage_gc = NULL; 1201 } 1202 1203 static int bch2_gc_done(struct bch_fs *c, 1204 bool initial, bool metadata_only) 1205 { 1206 struct bch_dev *ca = NULL; 1207 struct printbuf buf = PRINTBUF; 1208 bool verify = !metadata_only && 1209 !c->opts.reconstruct_alloc && 1210 (!initial || (c->sb.compat & (1ULL << BCH_COMPAT_alloc_info))); 1211 unsigned i; 1212 int ret = 0; 1213 1214 percpu_down_write(&c->mark_lock); 1215 1216 #define copy_field(_err, _f, _msg, ...) \ 1217 if (dst->_f != src->_f && \ 1218 (!verify || \ 1219 fsck_err(c, _err, _msg ": got %llu, should be %llu" \ 1220 , ##__VA_ARGS__, dst->_f, src->_f))) \ 1221 dst->_f = src->_f 1222 #define copy_dev_field(_err, _f, _msg, ...) \ 1223 copy_field(_err, _f, "dev %u has wrong " _msg, ca->dev_idx, ##__VA_ARGS__) 1224 #define copy_fs_field(_err, _f, _msg, ...) \ 1225 copy_field(_err, _f, "fs has wrong " _msg, ##__VA_ARGS__) 1226 1227 for (i = 0; i < ARRAY_SIZE(c->usage); i++) 1228 bch2_fs_usage_acc_to_base(c, i); 1229 1230 __for_each_member_device(c, ca) { 1231 struct bch_dev_usage *dst = ca->usage_base; 1232 struct bch_dev_usage *src = (void *) 1233 bch2_acc_percpu_u64s((u64 __percpu *) ca->usage_gc, 1234 dev_usage_u64s()); 1235 1236 for (i = 0; i < BCH_DATA_NR; i++) { 1237 copy_dev_field(dev_usage_buckets_wrong, 1238 d[i].buckets, "%s buckets", bch2_data_type_str(i)); 1239 copy_dev_field(dev_usage_sectors_wrong, 1240 d[i].sectors, "%s sectors", bch2_data_type_str(i)); 1241 copy_dev_field(dev_usage_fragmented_wrong, 1242 d[i].fragmented, "%s fragmented", bch2_data_type_str(i)); 1243 } 1244 } 1245 1246 { 1247 unsigned nr = fs_usage_u64s(c); 1248 struct bch_fs_usage *dst = c->usage_base; 1249 struct bch_fs_usage *src = (void *) 1250 bch2_acc_percpu_u64s((u64 __percpu *) c->usage_gc, nr); 1251 1252 copy_fs_field(fs_usage_hidden_wrong, 1253 b.hidden, "hidden"); 1254 copy_fs_field(fs_usage_btree_wrong, 1255 b.btree, "btree"); 1256 1257 if (!metadata_only) { 1258 copy_fs_field(fs_usage_data_wrong, 1259 b.data, "data"); 1260 copy_fs_field(fs_usage_cached_wrong, 1261 b.cached, "cached"); 1262 copy_fs_field(fs_usage_reserved_wrong, 1263 b.reserved, "reserved"); 1264 copy_fs_field(fs_usage_nr_inodes_wrong, 1265 b.nr_inodes,"nr_inodes"); 1266 1267 for (i = 0; i < BCH_REPLICAS_MAX; i++) 1268 copy_fs_field(fs_usage_persistent_reserved_wrong, 1269 persistent_reserved[i], 1270 "persistent_reserved[%i]", i); 1271 } 1272 1273 for (i = 0; i < c->replicas.nr; i++) { 1274 struct bch_replicas_entry_v1 *e = 1275 cpu_replicas_entry(&c->replicas, i); 1276 1277 if (metadata_only && 1278 (e->data_type == BCH_DATA_user || 1279 e->data_type == BCH_DATA_cached)) 1280 continue; 1281 1282 printbuf_reset(&buf); 1283 bch2_replicas_entry_to_text(&buf, e); 1284 1285 copy_fs_field(fs_usage_replicas_wrong, 1286 replicas[i], "%s", buf.buf); 1287 } 1288 } 1289 1290 #undef copy_fs_field 1291 #undef copy_dev_field 1292 #undef copy_stripe_field 1293 #undef copy_field 1294 fsck_err: 1295 if (ca) 1296 percpu_ref_put(&ca->ref); 1297 bch_err_fn(c, ret); 1298 1299 percpu_up_write(&c->mark_lock); 1300 printbuf_exit(&buf); 1301 return ret; 1302 } 1303 1304 static int bch2_gc_start(struct bch_fs *c) 1305 { 1306 BUG_ON(c->usage_gc); 1307 1308 c->usage_gc = __alloc_percpu_gfp(fs_usage_u64s(c) * sizeof(u64), 1309 sizeof(u64), GFP_KERNEL); 1310 if (!c->usage_gc) { 1311 bch_err(c, "error allocating c->usage_gc"); 1312 return -BCH_ERR_ENOMEM_gc_start; 1313 } 1314 1315 for_each_member_device(c, ca) { 1316 BUG_ON(ca->usage_gc); 1317 1318 ca->usage_gc = alloc_percpu(struct bch_dev_usage); 1319 if (!ca->usage_gc) { 1320 bch_err(c, "error allocating ca->usage_gc"); 1321 percpu_ref_put(&ca->ref); 1322 return -BCH_ERR_ENOMEM_gc_start; 1323 } 1324 1325 this_cpu_write(ca->usage_gc->d[BCH_DATA_free].buckets, 1326 ca->mi.nbuckets - ca->mi.first_bucket); 1327 } 1328 1329 return 0; 1330 } 1331 1332 static int bch2_gc_reset(struct bch_fs *c) 1333 { 1334 for_each_member_device(c, ca) { 1335 free_percpu(ca->usage_gc); 1336 ca->usage_gc = NULL; 1337 } 1338 1339 free_percpu(c->usage_gc); 1340 c->usage_gc = NULL; 1341 1342 return bch2_gc_start(c); 1343 } 1344 1345 /* returns true if not equal */ 1346 static inline bool bch2_alloc_v4_cmp(struct bch_alloc_v4 l, 1347 struct bch_alloc_v4 r) 1348 { 1349 return l.gen != r.gen || 1350 l.oldest_gen != r.oldest_gen || 1351 l.data_type != r.data_type || 1352 l.dirty_sectors != r.dirty_sectors || 1353 l.cached_sectors != r.cached_sectors || 1354 l.stripe_redundancy != r.stripe_redundancy || 1355 l.stripe != r.stripe; 1356 } 1357 1358 static int bch2_alloc_write_key(struct btree_trans *trans, 1359 struct btree_iter *iter, 1360 struct bkey_s_c k, 1361 bool metadata_only) 1362 { 1363 struct bch_fs *c = trans->c; 1364 struct bch_dev *ca = bch_dev_bkey_exists(c, iter->pos.inode); 1365 struct bucket old_gc, gc, *b; 1366 struct bkey_i_alloc_v4 *a; 1367 struct bch_alloc_v4 old_convert, new; 1368 const struct bch_alloc_v4 *old; 1369 int ret; 1370 1371 old = bch2_alloc_to_v4(k, &old_convert); 1372 new = *old; 1373 1374 percpu_down_read(&c->mark_lock); 1375 b = gc_bucket(ca, iter->pos.offset); 1376 old_gc = *b; 1377 1378 if ((old->data_type == BCH_DATA_sb || 1379 old->data_type == BCH_DATA_journal) && 1380 !bch2_dev_is_online(ca)) { 1381 b->data_type = old->data_type; 1382 b->dirty_sectors = old->dirty_sectors; 1383 } 1384 1385 /* 1386 * b->data_type doesn't yet include need_discard & need_gc_gen states - 1387 * fix that here: 1388 */ 1389 b->data_type = __alloc_data_type(b->dirty_sectors, 1390 b->cached_sectors, 1391 b->stripe, 1392 *old, 1393 b->data_type); 1394 gc = *b; 1395 1396 if (gc.data_type != old_gc.data_type || 1397 gc.dirty_sectors != old_gc.dirty_sectors) 1398 bch2_dev_usage_update_m(c, ca, &old_gc, &gc); 1399 percpu_up_read(&c->mark_lock); 1400 1401 if (metadata_only && 1402 gc.data_type != BCH_DATA_sb && 1403 gc.data_type != BCH_DATA_journal && 1404 gc.data_type != BCH_DATA_btree) 1405 return 0; 1406 1407 if (gen_after(old->gen, gc.gen)) 1408 return 0; 1409 1410 if (fsck_err_on(new.data_type != gc.data_type, c, 1411 alloc_key_data_type_wrong, 1412 "bucket %llu:%llu gen %u has wrong data_type" 1413 ": got %s, should be %s", 1414 iter->pos.inode, iter->pos.offset, 1415 gc.gen, 1416 bch2_data_type_str(new.data_type), 1417 bch2_data_type_str(gc.data_type))) 1418 new.data_type = gc.data_type; 1419 1420 #define copy_bucket_field(_errtype, _f) \ 1421 if (fsck_err_on(new._f != gc._f, c, _errtype, \ 1422 "bucket %llu:%llu gen %u data type %s has wrong " #_f \ 1423 ": got %u, should be %u", \ 1424 iter->pos.inode, iter->pos.offset, \ 1425 gc.gen, \ 1426 bch2_data_type_str(gc.data_type), \ 1427 new._f, gc._f)) \ 1428 new._f = gc._f; \ 1429 1430 copy_bucket_field(alloc_key_gen_wrong, 1431 gen); 1432 copy_bucket_field(alloc_key_dirty_sectors_wrong, 1433 dirty_sectors); 1434 copy_bucket_field(alloc_key_cached_sectors_wrong, 1435 cached_sectors); 1436 copy_bucket_field(alloc_key_stripe_wrong, 1437 stripe); 1438 copy_bucket_field(alloc_key_stripe_redundancy_wrong, 1439 stripe_redundancy); 1440 #undef copy_bucket_field 1441 1442 if (!bch2_alloc_v4_cmp(*old, new)) 1443 return 0; 1444 1445 a = bch2_alloc_to_v4_mut(trans, k); 1446 ret = PTR_ERR_OR_ZERO(a); 1447 if (ret) 1448 return ret; 1449 1450 a->v = new; 1451 1452 /* 1453 * The trigger normally makes sure this is set, but we're not running 1454 * triggers: 1455 */ 1456 if (a->v.data_type == BCH_DATA_cached && !a->v.io_time[READ]) 1457 a->v.io_time[READ] = max_t(u64, 1, atomic64_read(&c->io_clock[READ].now)); 1458 1459 ret = bch2_trans_update(trans, iter, &a->k_i, BTREE_TRIGGER_NORUN); 1460 fsck_err: 1461 return ret; 1462 } 1463 1464 static int bch2_gc_alloc_done(struct bch_fs *c, bool metadata_only) 1465 { 1466 int ret = 0; 1467 1468 for_each_member_device(c, ca) { 1469 ret = bch2_trans_run(c, 1470 for_each_btree_key_upto_commit(trans, iter, BTREE_ID_alloc, 1471 POS(ca->dev_idx, ca->mi.first_bucket), 1472 POS(ca->dev_idx, ca->mi.nbuckets - 1), 1473 BTREE_ITER_SLOTS|BTREE_ITER_PREFETCH, k, 1474 NULL, NULL, BCH_TRANS_COMMIT_lazy_rw, 1475 bch2_alloc_write_key(trans, &iter, k, metadata_only))); 1476 if (ret) { 1477 percpu_ref_put(&ca->ref); 1478 break; 1479 } 1480 } 1481 1482 bch_err_fn(c, ret); 1483 return ret; 1484 } 1485 1486 static int bch2_gc_alloc_start(struct bch_fs *c, bool metadata_only) 1487 { 1488 for_each_member_device(c, ca) { 1489 struct bucket_array *buckets = kvmalloc(sizeof(struct bucket_array) + 1490 ca->mi.nbuckets * sizeof(struct bucket), 1491 GFP_KERNEL|__GFP_ZERO); 1492 if (!buckets) { 1493 percpu_ref_put(&ca->ref); 1494 bch_err(c, "error allocating ca->buckets[gc]"); 1495 return -BCH_ERR_ENOMEM_gc_alloc_start; 1496 } 1497 1498 buckets->first_bucket = ca->mi.first_bucket; 1499 buckets->nbuckets = ca->mi.nbuckets; 1500 rcu_assign_pointer(ca->buckets_gc, buckets); 1501 } 1502 1503 int ret = bch2_trans_run(c, 1504 for_each_btree_key(trans, iter, BTREE_ID_alloc, POS_MIN, 1505 BTREE_ITER_PREFETCH, k, ({ 1506 struct bch_dev *ca = bch_dev_bkey_exists(c, k.k->p.inode); 1507 struct bucket *g = gc_bucket(ca, k.k->p.offset); 1508 1509 struct bch_alloc_v4 a_convert; 1510 const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert); 1511 1512 g->gen_valid = 1; 1513 g->gen = a->gen; 1514 1515 if (metadata_only && 1516 (a->data_type == BCH_DATA_user || 1517 a->data_type == BCH_DATA_cached || 1518 a->data_type == BCH_DATA_parity)) { 1519 g->data_type = a->data_type; 1520 g->dirty_sectors = a->dirty_sectors; 1521 g->cached_sectors = a->cached_sectors; 1522 g->stripe = a->stripe; 1523 g->stripe_redundancy = a->stripe_redundancy; 1524 } 1525 1526 0; 1527 }))); 1528 bch_err_fn(c, ret); 1529 return ret; 1530 } 1531 1532 static void bch2_gc_alloc_reset(struct bch_fs *c, bool metadata_only) 1533 { 1534 for_each_member_device(c, ca) { 1535 struct bucket_array *buckets = gc_bucket_array(ca); 1536 struct bucket *g; 1537 1538 for_each_bucket(g, buckets) { 1539 if (metadata_only && 1540 (g->data_type == BCH_DATA_user || 1541 g->data_type == BCH_DATA_cached || 1542 g->data_type == BCH_DATA_parity)) 1543 continue; 1544 g->data_type = 0; 1545 g->dirty_sectors = 0; 1546 g->cached_sectors = 0; 1547 } 1548 } 1549 } 1550 1551 static int bch2_gc_write_reflink_key(struct btree_trans *trans, 1552 struct btree_iter *iter, 1553 struct bkey_s_c k, 1554 size_t *idx) 1555 { 1556 struct bch_fs *c = trans->c; 1557 const __le64 *refcount = bkey_refcount_c(k); 1558 struct printbuf buf = PRINTBUF; 1559 struct reflink_gc *r; 1560 int ret = 0; 1561 1562 if (!refcount) 1563 return 0; 1564 1565 while ((r = genradix_ptr(&c->reflink_gc_table, *idx)) && 1566 r->offset < k.k->p.offset) 1567 ++*idx; 1568 1569 if (!r || 1570 r->offset != k.k->p.offset || 1571 r->size != k.k->size) { 1572 bch_err(c, "unexpected inconsistency walking reflink table at gc finish"); 1573 return -EINVAL; 1574 } 1575 1576 if (fsck_err_on(r->refcount != le64_to_cpu(*refcount), c, 1577 reflink_v_refcount_wrong, 1578 "reflink key has wrong refcount:\n" 1579 " %s\n" 1580 " should be %u", 1581 (bch2_bkey_val_to_text(&buf, c, k), buf.buf), 1582 r->refcount)) { 1583 struct bkey_i *new = bch2_bkey_make_mut_noupdate(trans, k); 1584 ret = PTR_ERR_OR_ZERO(new); 1585 if (ret) 1586 return ret; 1587 1588 if (!r->refcount) 1589 new->k.type = KEY_TYPE_deleted; 1590 else 1591 *bkey_refcount(bkey_i_to_s(new)) = cpu_to_le64(r->refcount); 1592 ret = bch2_trans_update(trans, iter, new, 0); 1593 } 1594 fsck_err: 1595 printbuf_exit(&buf); 1596 return ret; 1597 } 1598 1599 static int bch2_gc_reflink_done(struct bch_fs *c, bool metadata_only) 1600 { 1601 size_t idx = 0; 1602 1603 if (metadata_only) 1604 return 0; 1605 1606 int ret = bch2_trans_run(c, 1607 for_each_btree_key_commit(trans, iter, 1608 BTREE_ID_reflink, POS_MIN, 1609 BTREE_ITER_PREFETCH, k, 1610 NULL, NULL, BCH_TRANS_COMMIT_no_enospc, 1611 bch2_gc_write_reflink_key(trans, &iter, k, &idx))); 1612 c->reflink_gc_nr = 0; 1613 return ret; 1614 } 1615 1616 static int bch2_gc_reflink_start(struct bch_fs *c, 1617 bool metadata_only) 1618 { 1619 1620 if (metadata_only) 1621 return 0; 1622 1623 c->reflink_gc_nr = 0; 1624 1625 int ret = bch2_trans_run(c, 1626 for_each_btree_key(trans, iter, BTREE_ID_reflink, POS_MIN, 1627 BTREE_ITER_PREFETCH, k, ({ 1628 const __le64 *refcount = bkey_refcount_c(k); 1629 1630 if (!refcount) 1631 continue; 1632 1633 struct reflink_gc *r = genradix_ptr_alloc(&c->reflink_gc_table, 1634 c->reflink_gc_nr++, GFP_KERNEL); 1635 if (!r) { 1636 ret = -BCH_ERR_ENOMEM_gc_reflink_start; 1637 break; 1638 } 1639 1640 r->offset = k.k->p.offset; 1641 r->size = k.k->size; 1642 r->refcount = 0; 1643 0; 1644 }))); 1645 1646 bch_err_fn(c, ret); 1647 return ret; 1648 } 1649 1650 static void bch2_gc_reflink_reset(struct bch_fs *c, bool metadata_only) 1651 { 1652 struct genradix_iter iter; 1653 struct reflink_gc *r; 1654 1655 genradix_for_each(&c->reflink_gc_table, iter, r) 1656 r->refcount = 0; 1657 } 1658 1659 static int bch2_gc_write_stripes_key(struct btree_trans *trans, 1660 struct btree_iter *iter, 1661 struct bkey_s_c k) 1662 { 1663 struct bch_fs *c = trans->c; 1664 struct printbuf buf = PRINTBUF; 1665 const struct bch_stripe *s; 1666 struct gc_stripe *m; 1667 bool bad = false; 1668 unsigned i; 1669 int ret = 0; 1670 1671 if (k.k->type != KEY_TYPE_stripe) 1672 return 0; 1673 1674 s = bkey_s_c_to_stripe(k).v; 1675 m = genradix_ptr(&c->gc_stripes, k.k->p.offset); 1676 1677 for (i = 0; i < s->nr_blocks; i++) { 1678 u32 old = stripe_blockcount_get(s, i); 1679 u32 new = (m ? m->block_sectors[i] : 0); 1680 1681 if (old != new) { 1682 prt_printf(&buf, "stripe block %u has wrong sector count: got %u, should be %u\n", 1683 i, old, new); 1684 bad = true; 1685 } 1686 } 1687 1688 if (bad) 1689 bch2_bkey_val_to_text(&buf, c, k); 1690 1691 if (fsck_err_on(bad, c, stripe_sector_count_wrong, 1692 "%s", buf.buf)) { 1693 struct bkey_i_stripe *new; 1694 1695 new = bch2_trans_kmalloc(trans, bkey_bytes(k.k)); 1696 ret = PTR_ERR_OR_ZERO(new); 1697 if (ret) 1698 return ret; 1699 1700 bkey_reassemble(&new->k_i, k); 1701 1702 for (i = 0; i < new->v.nr_blocks; i++) 1703 stripe_blockcount_set(&new->v, i, m ? m->block_sectors[i] : 0); 1704 1705 ret = bch2_trans_update(trans, iter, &new->k_i, 0); 1706 } 1707 fsck_err: 1708 printbuf_exit(&buf); 1709 return ret; 1710 } 1711 1712 static int bch2_gc_stripes_done(struct bch_fs *c, bool metadata_only) 1713 { 1714 if (metadata_only) 1715 return 0; 1716 1717 return bch2_trans_run(c, 1718 for_each_btree_key_commit(trans, iter, 1719 BTREE_ID_stripes, POS_MIN, 1720 BTREE_ITER_PREFETCH, k, 1721 NULL, NULL, BCH_TRANS_COMMIT_no_enospc, 1722 bch2_gc_write_stripes_key(trans, &iter, k))); 1723 } 1724 1725 static void bch2_gc_stripes_reset(struct bch_fs *c, bool metadata_only) 1726 { 1727 genradix_free(&c->gc_stripes); 1728 } 1729 1730 /** 1731 * bch2_gc - walk _all_ references to buckets, and recompute them: 1732 * 1733 * @c: filesystem object 1734 * @initial: are we in recovery? 1735 * @metadata_only: are we just checking metadata references, or everything? 1736 * 1737 * Returns: 0 on success, or standard errcode on failure 1738 * 1739 * Order matters here: 1740 * - Concurrent GC relies on the fact that we have a total ordering for 1741 * everything that GC walks - see gc_will_visit_node(), 1742 * gc_will_visit_root() 1743 * 1744 * - also, references move around in the course of index updates and 1745 * various other crap: everything needs to agree on the ordering 1746 * references are allowed to move around in - e.g., we're allowed to 1747 * start with a reference owned by an open_bucket (the allocator) and 1748 * move it to the btree, but not the reverse. 1749 * 1750 * This is necessary to ensure that gc doesn't miss references that 1751 * move around - if references move backwards in the ordering GC 1752 * uses, GC could skip past them 1753 */ 1754 int bch2_gc(struct bch_fs *c, bool initial, bool metadata_only) 1755 { 1756 unsigned iter = 0; 1757 int ret; 1758 1759 lockdep_assert_held(&c->state_lock); 1760 1761 down_write(&c->gc_lock); 1762 1763 bch2_btree_interior_updates_flush(c); 1764 1765 ret = bch2_gc_start(c) ?: 1766 bch2_gc_alloc_start(c, metadata_only) ?: 1767 bch2_gc_reflink_start(c, metadata_only); 1768 if (ret) 1769 goto out; 1770 again: 1771 gc_pos_set(c, gc_phase(GC_PHASE_START)); 1772 1773 bch2_mark_superblocks(c); 1774 1775 ret = bch2_gc_btrees(c, initial, metadata_only); 1776 1777 if (ret) 1778 goto out; 1779 1780 #if 0 1781 bch2_mark_pending_btree_node_frees(c); 1782 #endif 1783 c->gc_count++; 1784 1785 if (test_bit(BCH_FS_need_another_gc, &c->flags) || 1786 (!iter && bch2_test_restart_gc)) { 1787 if (iter++ > 2) { 1788 bch_info(c, "Unable to fix bucket gens, looping"); 1789 ret = -EINVAL; 1790 goto out; 1791 } 1792 1793 /* 1794 * XXX: make sure gens we fixed got saved 1795 */ 1796 bch_info(c, "Second GC pass needed, restarting:"); 1797 clear_bit(BCH_FS_need_another_gc, &c->flags); 1798 __gc_pos_set(c, gc_phase(GC_PHASE_NOT_RUNNING)); 1799 1800 bch2_gc_stripes_reset(c, metadata_only); 1801 bch2_gc_alloc_reset(c, metadata_only); 1802 bch2_gc_reflink_reset(c, metadata_only); 1803 ret = bch2_gc_reset(c); 1804 if (ret) 1805 goto out; 1806 1807 /* flush fsck errors, reset counters */ 1808 bch2_flush_fsck_errs(c); 1809 goto again; 1810 } 1811 out: 1812 if (!ret) { 1813 bch2_journal_block(&c->journal); 1814 1815 ret = bch2_gc_alloc_done(c, metadata_only) ?: 1816 bch2_gc_done(c, initial, metadata_only) ?: 1817 bch2_gc_stripes_done(c, metadata_only) ?: 1818 bch2_gc_reflink_done(c, metadata_only); 1819 1820 bch2_journal_unblock(&c->journal); 1821 } 1822 1823 percpu_down_write(&c->mark_lock); 1824 /* Indicates that gc is no longer in progress: */ 1825 __gc_pos_set(c, gc_phase(GC_PHASE_NOT_RUNNING)); 1826 1827 bch2_gc_free(c); 1828 percpu_up_write(&c->mark_lock); 1829 1830 up_write(&c->gc_lock); 1831 1832 /* 1833 * At startup, allocations can happen directly instead of via the 1834 * allocator thread - issue wakeup in case they blocked on gc_lock: 1835 */ 1836 closure_wake_up(&c->freelist_wait); 1837 bch_err_fn(c, ret); 1838 return ret; 1839 } 1840 1841 static int gc_btree_gens_key(struct btree_trans *trans, 1842 struct btree_iter *iter, 1843 struct bkey_s_c k) 1844 { 1845 struct bch_fs *c = trans->c; 1846 struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); 1847 struct bkey_i *u; 1848 int ret; 1849 1850 percpu_down_read(&c->mark_lock); 1851 bkey_for_each_ptr(ptrs, ptr) { 1852 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 1853 1854 if (ptr_stale(ca, ptr) > 16) { 1855 percpu_up_read(&c->mark_lock); 1856 goto update; 1857 } 1858 } 1859 1860 bkey_for_each_ptr(ptrs, ptr) { 1861 struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev); 1862 u8 *gen = &ca->oldest_gen[PTR_BUCKET_NR(ca, ptr)]; 1863 1864 if (gen_after(*gen, ptr->gen)) 1865 *gen = ptr->gen; 1866 } 1867 percpu_up_read(&c->mark_lock); 1868 return 0; 1869 update: 1870 u = bch2_bkey_make_mut(trans, iter, &k, 0); 1871 ret = PTR_ERR_OR_ZERO(u); 1872 if (ret) 1873 return ret; 1874 1875 bch2_extent_normalize(c, bkey_i_to_s(u)); 1876 return 0; 1877 } 1878 1879 static int bch2_alloc_write_oldest_gen(struct btree_trans *trans, struct btree_iter *iter, 1880 struct bkey_s_c k) 1881 { 1882 struct bch_dev *ca = bch_dev_bkey_exists(trans->c, iter->pos.inode); 1883 struct bch_alloc_v4 a_convert; 1884 const struct bch_alloc_v4 *a = bch2_alloc_to_v4(k, &a_convert); 1885 struct bkey_i_alloc_v4 *a_mut; 1886 int ret; 1887 1888 if (a->oldest_gen == ca->oldest_gen[iter->pos.offset]) 1889 return 0; 1890 1891 a_mut = bch2_alloc_to_v4_mut(trans, k); 1892 ret = PTR_ERR_OR_ZERO(a_mut); 1893 if (ret) 1894 return ret; 1895 1896 a_mut->v.oldest_gen = ca->oldest_gen[iter->pos.offset]; 1897 a_mut->v.data_type = alloc_data_type(a_mut->v, a_mut->v.data_type); 1898 1899 return bch2_trans_update(trans, iter, &a_mut->k_i, 0); 1900 } 1901 1902 int bch2_gc_gens(struct bch_fs *c) 1903 { 1904 u64 b, start_time = local_clock(); 1905 int ret; 1906 1907 /* 1908 * Ideally we would be using state_lock and not gc_lock here, but that 1909 * introduces a deadlock in the RO path - we currently take the state 1910 * lock at the start of going RO, thus the gc thread may get stuck: 1911 */ 1912 if (!mutex_trylock(&c->gc_gens_lock)) 1913 return 0; 1914 1915 trace_and_count(c, gc_gens_start, c); 1916 down_read(&c->gc_lock); 1917 1918 for_each_member_device(c, ca) { 1919 struct bucket_gens *gens = bucket_gens(ca); 1920 1921 BUG_ON(ca->oldest_gen); 1922 1923 ca->oldest_gen = kvmalloc(gens->nbuckets, GFP_KERNEL); 1924 if (!ca->oldest_gen) { 1925 percpu_ref_put(&ca->ref); 1926 ret = -BCH_ERR_ENOMEM_gc_gens; 1927 goto err; 1928 } 1929 1930 for (b = gens->first_bucket; 1931 b < gens->nbuckets; b++) 1932 ca->oldest_gen[b] = gens->b[b]; 1933 } 1934 1935 for (unsigned i = 0; i < BTREE_ID_NR; i++) 1936 if (btree_type_has_ptrs(i)) { 1937 c->gc_gens_btree = i; 1938 c->gc_gens_pos = POS_MIN; 1939 1940 ret = bch2_trans_run(c, 1941 for_each_btree_key_commit(trans, iter, i, 1942 POS_MIN, 1943 BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, 1944 k, 1945 NULL, NULL, 1946 BCH_TRANS_COMMIT_no_enospc, 1947 gc_btree_gens_key(trans, &iter, k))); 1948 if (ret) 1949 goto err; 1950 } 1951 1952 ret = bch2_trans_run(c, 1953 for_each_btree_key_commit(trans, iter, BTREE_ID_alloc, 1954 POS_MIN, 1955 BTREE_ITER_PREFETCH, 1956 k, 1957 NULL, NULL, 1958 BCH_TRANS_COMMIT_no_enospc, 1959 bch2_alloc_write_oldest_gen(trans, &iter, k))); 1960 if (ret) 1961 goto err; 1962 1963 c->gc_gens_btree = 0; 1964 c->gc_gens_pos = POS_MIN; 1965 1966 c->gc_count++; 1967 1968 bch2_time_stats_update(&c->times[BCH_TIME_btree_gc], start_time); 1969 trace_and_count(c, gc_gens_end, c); 1970 err: 1971 for_each_member_device(c, ca) { 1972 kvfree(ca->oldest_gen); 1973 ca->oldest_gen = NULL; 1974 } 1975 1976 up_read(&c->gc_lock); 1977 mutex_unlock(&c->gc_gens_lock); 1978 if (!bch2_err_matches(ret, EROFS)) 1979 bch_err_fn(c, ret); 1980 return ret; 1981 } 1982 1983 static int bch2_gc_thread(void *arg) 1984 { 1985 struct bch_fs *c = arg; 1986 struct io_clock *clock = &c->io_clock[WRITE]; 1987 unsigned long last = atomic64_read(&clock->now); 1988 unsigned last_kick = atomic_read(&c->kick_gc); 1989 1990 set_freezable(); 1991 1992 while (1) { 1993 while (1) { 1994 set_current_state(TASK_INTERRUPTIBLE); 1995 1996 if (kthread_should_stop()) { 1997 __set_current_state(TASK_RUNNING); 1998 return 0; 1999 } 2000 2001 if (atomic_read(&c->kick_gc) != last_kick) 2002 break; 2003 2004 if (c->btree_gc_periodic) { 2005 unsigned long next = last + c->capacity / 16; 2006 2007 if (atomic64_read(&clock->now) >= next) 2008 break; 2009 2010 bch2_io_clock_schedule_timeout(clock, next); 2011 } else { 2012 schedule(); 2013 } 2014 2015 try_to_freeze(); 2016 } 2017 __set_current_state(TASK_RUNNING); 2018 2019 last = atomic64_read(&clock->now); 2020 last_kick = atomic_read(&c->kick_gc); 2021 2022 /* 2023 * Full gc is currently incompatible with btree key cache: 2024 */ 2025 #if 0 2026 ret = bch2_gc(c, false, false); 2027 #else 2028 bch2_gc_gens(c); 2029 #endif 2030 debug_check_no_locks_held(); 2031 } 2032 2033 return 0; 2034 } 2035 2036 void bch2_gc_thread_stop(struct bch_fs *c) 2037 { 2038 struct task_struct *p; 2039 2040 p = c->gc_thread; 2041 c->gc_thread = NULL; 2042 2043 if (p) { 2044 kthread_stop(p); 2045 put_task_struct(p); 2046 } 2047 } 2048 2049 int bch2_gc_thread_start(struct bch_fs *c) 2050 { 2051 struct task_struct *p; 2052 2053 if (c->gc_thread) 2054 return 0; 2055 2056 p = kthread_create(bch2_gc_thread, c, "bch-gc/%s", c->name); 2057 if (IS_ERR(p)) { 2058 bch_err_fn(c, PTR_ERR(p)); 2059 return PTR_ERR(p); 2060 } 2061 2062 get_task_struct(p); 2063 c->gc_thread = p; 2064 wake_up_process(p); 2065 return 0; 2066 } 2067