X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Finode.c;h=8bfd99cb7ad149deca66f2998bcecbc4bcc076a1;hb=1e358401ecdf1963e5799de19ab69111e82e5ebc;hp=938c7b43d0b468dab3c1fc1a8eb7804d6f1d1599;hpb=5a5a6c25a93e83e0b1e659d7808e7636d0b770ed;p=bcachefs-tools-debian diff --git a/libbcachefs/inode.c b/libbcachefs/inode.c index 938c7b4..8bfd99c 100644 --- a/libbcachefs/inode.c +++ b/libbcachefs/inode.c @@ -2,6 +2,7 @@ #include "bcachefs.h" #include "btree_key_cache.h" +#include "btree_write_buffer.h" #include "bkey_methods.h" #include "btree_update.h" #include "buckets.h" @@ -10,6 +11,7 @@ #include "extent_update.h" #include "inode.h" #include "str_hash.h" +#include "snapshot.h" #include "subvolume.h" #include "varint.h" @@ -118,8 +120,7 @@ static inline void bch2_inode_pack_inlined(struct bkey_inode_buf *packed, if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG)) { struct bch_inode_unpacked unpacked; - int ret = bch2_inode_unpack(bkey_i_to_s_c(&packed->inode.k_i), - &unpacked); + ret = bch2_inode_unpack(bkey_i_to_s_c(&packed->inode.k_i), &unpacked); BUG_ON(ret); BUG_ON(unpacked.bi_inum != inode->bi_inum); BUG_ON(unpacked.bi_hash_seed != inode->bi_hash_seed); @@ -269,6 +270,8 @@ static int bch2_inode_unpack_v3(struct bkey_s_c k, static noinline int bch2_inode_unpack_slowpath(struct bkey_s_c k, struct bch_inode_unpacked *unpacked) { + memset(unpacked, 0, sizeof(*unpacked)); + switch (k.k->type) { case KEY_TYPE_inode: { struct bkey_s_c_inode inode = bkey_s_c_to_inode(k); @@ -314,7 +317,7 @@ int bch2_inode_unpack(struct bkey_s_c k, return bch2_inode_unpack_slowpath(k, unpacked); } -int bch2_inode_peek(struct btree_trans *trans, +static int bch2_inode_peek_nowarn(struct btree_trans *trans, struct btree_iter *iter, struct bch_inode_unpacked *inode, subvol_inum inum, unsigned flags) @@ -327,15 +330,14 @@ int bch2_inode_peek(struct btree_trans *trans, if (ret) return ret; - bch2_trans_iter_init(trans, iter, BTREE_ID_inodes, - SPOS(0, inum.inum, snapshot), - flags|BTREE_ITER_CACHED); - k = bch2_btree_iter_peek_slot(iter); + k = bch2_bkey_get_iter(trans, iter, BTREE_ID_inodes, + SPOS(0, inum.inum, snapshot), + flags|BTREE_ITER_CACHED); ret = bkey_err(k); if (ret) - goto err; + return ret; - ret = bkey_is_inode(k.k) ? 0 : -ENOENT; + ret = bkey_is_inode(k.k) ? 0 : -BCH_ERR_ENOENT_inode; if (ret) goto err; @@ -349,6 +351,18 @@ err: return ret; } +int bch2_inode_peek(struct btree_trans *trans, + struct btree_iter *iter, + struct bch_inode_unpacked *inode, + subvol_inum inum, unsigned flags) +{ + int ret = bch2_inode_peek_nowarn(trans, iter, inode, inum, flags); + + if (ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart)) + bch_err_msg(trans->c, ret, "looking up inum %u:%llu:", inum.subvol, inum.inum); + return ret; +} + int bch2_inode_write(struct btree_trans *trans, struct btree_iter *iter, struct bch_inode_unpacked *inode) @@ -431,16 +445,11 @@ static int __bch2_inode_invalid(struct bkey_s_c k, struct printbuf *err) } int bch2_inode_invalid(const struct bch_fs *c, struct bkey_s_c k, - int rw, struct printbuf *err) + enum bkey_invalid_flags flags, + struct printbuf *err) { struct bkey_s_c_inode inode = bkey_s_c_to_inode(k); - if (bkey_val_bytes(k.k) < sizeof(*inode.v)) { - prt_printf(err, "incorrect value size (%zu < %zu)", - bkey_val_bytes(k.k), sizeof(*inode.v)); - return -BCH_ERR_invalid_bkey; - } - if (INODE_STR_HASH(inode.v) >= BCH_STR_HASH_NR) { prt_printf(err, "invalid str hash type (%llu >= %u)", INODE_STR_HASH(inode.v), BCH_STR_HASH_NR); @@ -451,16 +460,11 @@ int bch2_inode_invalid(const struct bch_fs *c, struct bkey_s_c k, } int bch2_inode_v2_invalid(const struct bch_fs *c, struct bkey_s_c k, - int rw, struct printbuf *err) + enum bkey_invalid_flags flags, + struct printbuf *err) { struct bkey_s_c_inode_v2 inode = bkey_s_c_to_inode_v2(k); - if (bkey_val_bytes(k.k) < sizeof(*inode.v)) { - prt_printf(err, "incorrect value size (%zu < %zu)", - bkey_val_bytes(k.k), sizeof(*inode.v)); - return -BCH_ERR_invalid_bkey; - } - if (INODEv2_STR_HASH(inode.v) >= BCH_STR_HASH_NR) { prt_printf(err, "invalid str hash type (%llu >= %u)", INODEv2_STR_HASH(inode.v), BCH_STR_HASH_NR); @@ -471,16 +475,11 @@ int bch2_inode_v2_invalid(const struct bch_fs *c, struct bkey_s_c k, } int bch2_inode_v3_invalid(const struct bch_fs *c, struct bkey_s_c k, - int rw, struct printbuf *err) + enum bkey_invalid_flags flags, + struct printbuf *err) { struct bkey_s_c_inode_v3 inode = bkey_s_c_to_inode_v3(k); - if (bkey_val_bytes(k.k) < sizeof(*inode.v)) { - prt_printf(err, "incorrect value size (%zu < %zu)", - bkey_val_bytes(k.k), sizeof(*inode.v)); - return -BCH_ERR_invalid_bkey; - } - if (INODEv3_FIELDS_START(inode.v) < INODEv3_FIELDS_START_INITIAL || INODEv3_FIELDS_START(inode.v) > bkey_val_u64s(inode.k)) { prt_printf(err, "invalid fields_start (got %llu, min %u max %zu)", @@ -533,20 +532,95 @@ void bch2_inode_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c __bch2_inode_unpacked_to_text(out, &inode); } +static inline u64 bkey_inode_flags(struct bkey_s_c k) +{ + switch (k.k->type) { + case KEY_TYPE_inode: + return le32_to_cpu(bkey_s_c_to_inode(k).v->bi_flags); + case KEY_TYPE_inode_v2: + return le64_to_cpu(bkey_s_c_to_inode_v2(k).v->bi_flags); + case KEY_TYPE_inode_v3: + return le64_to_cpu(bkey_s_c_to_inode_v3(k).v->bi_flags); + default: + return 0; + } +} + +static inline bool bkey_is_deleted_inode(struct bkey_s_c k) +{ + return bkey_inode_flags(k) & BCH_INODE_UNLINKED; +} + +int bch2_trans_mark_inode(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, + struct bkey_i *new, + unsigned flags) +{ + int nr = bkey_is_inode(&new->k) - bkey_is_inode(old.k); + bool old_deleted = bkey_is_deleted_inode(old); + bool new_deleted = bkey_is_deleted_inode(bkey_i_to_s_c(new)); + + if (nr) { + int ret = bch2_replicas_deltas_realloc(trans, 0); + struct replicas_delta_list *d = trans->fs_usage_deltas; + + if (ret) + return ret; + + d->nr_inodes += nr; + } + + if (old_deleted != new_deleted) { + int ret = bch2_btree_bit_mod(trans, BTREE_ID_deleted_inodes, new->k.p, new_deleted); + if (ret) + return ret; + } + + return 0; +} + +int bch2_mark_inode(struct btree_trans *trans, + enum btree_id btree_id, unsigned level, + struct bkey_s_c old, struct bkey_s_c new, + unsigned flags) +{ + struct bch_fs *c = trans->c; + struct bch_fs_usage *fs_usage; + u64 journal_seq = trans->journal_res.seq; + + if (flags & BTREE_TRIGGER_INSERT) { + struct bch_inode_v3 *v = (struct bch_inode_v3 *) new.v; + + BUG_ON(!journal_seq); + BUG_ON(new.k->type != KEY_TYPE_inode_v3); + + v->bi_journal_seq = cpu_to_le64(journal_seq); + } + + if (flags & BTREE_TRIGGER_GC) { + percpu_down_read(&c->mark_lock); + preempt_disable(); + + fs_usage = fs_usage_ptr(c, journal_seq, flags & BTREE_TRIGGER_GC); + fs_usage->nr_inodes += bkey_is_inode(new.k); + fs_usage->nr_inodes -= bkey_is_inode(old.k); + + preempt_enable(); + percpu_up_read(&c->mark_lock); + } + return 0; +} + int bch2_inode_generation_invalid(const struct bch_fs *c, struct bkey_s_c k, - int rw, struct printbuf *err) + enum bkey_invalid_flags flags, + struct printbuf *err) { if (k.k->p.inode) { prt_printf(err, "nonzero k.p.inode"); return -BCH_ERR_invalid_bkey; } - if (bkey_val_bytes(k.k) != sizeof(struct bch_inode_generation)) { - prt_printf(err, "incorrect value size (%zu != %zu)", - bkey_val_bytes(k.k), sizeof(struct bch_inode_generation)); - return -BCH_ERR_invalid_bkey; - } - return 0; } @@ -661,19 +735,8 @@ again: while ((k = bch2_btree_iter_peek(iter)).k && !(ret = bkey_err(k)) && bkey_lt(k.k->p, POS(0, max))) { - while (pos < iter->pos.offset) { - if (!bch2_btree_key_cache_find(c, BTREE_ID_inodes, POS(0, pos))) - goto found_slot; - - pos++; - } - - if (k.k->p.snapshot == snapshot && - !bkey_is_inode(k.k) && - !bch2_btree_key_cache_find(c, BTREE_ID_inodes, SPOS(0, pos, snapshot))) { - bch2_btree_iter_advance(iter); - continue; - } + if (pos < iter->pos.offset) + goto found_slot; /* * We don't need to iterate over keys in every snapshot once @@ -683,12 +746,8 @@ again: bch2_btree_iter_set_pos(iter, POS(0, pos)); } - while (!ret && pos < max) { - if (!bch2_btree_key_cache_find(c, BTREE_ID_inodes, POS(0, pos))) - goto found_slot; - - pos++; - } + if (!ret && pos < max) + goto found_slot; if (!ret && start == min) ret = -BCH_ERR_ENOSPC_inode_create; @@ -711,11 +770,6 @@ found_slot: return ret; } - /* We may have raced while the iterator wasn't pointing at pos: */ - if (bkey_is_inode(k.k) || - bch2_btree_key_cache_find(c, BTREE_ID_inodes, k.k->p)) - goto again; - *hint = k.k->p.offset; inode_u->bi_inum = k.k->p.offset; inode_u->bi_generation = bkey_generation(k); @@ -732,11 +786,11 @@ static int bch2_inode_delete_keys(struct btree_trans *trans, int ret = 0; /* - * We're never going to be deleting extents, no need to use an extent - * iterator: + * We're never going to be deleting partial extents, no need to use an + * extent iterator: */ bch2_trans_iter_init(trans, &iter, id, POS(inum.inum, 0), - BTREE_ITER_INTENT); + BTREE_ITER_INTENT|BTREE_ITER_NOT_EXTENTS); while (1) { bch2_trans_begin(trans); @@ -758,14 +812,6 @@ static int bch2_inode_delete_keys(struct btree_trans *trans, bkey_init(&delete.k); delete.k.p = iter.pos; - if (iter.flags & BTREE_ITER_IS_EXTENTS) { - bch2_key_resize(&delete.k, k.k->p.offset - iter.pos.offset); - - ret = bch2_extent_trim_atomic(trans, &iter, &delete); - if (ret) - goto err; - } - ret = bch2_trans_update(trans, &iter, &delete, 0) ?: bch2_trans_commit(trans, NULL, NULL, BTREE_INSERT_NOFAIL); @@ -780,7 +826,7 @@ err: int bch2_inode_rm(struct bch_fs *c, subvol_inum inum) { - struct btree_trans trans; + struct btree_trans *trans = bch2_trans_get(c); struct btree_iter iter = { NULL }; struct bkey_i_inode_generation delete; struct bch_inode_unpacked inode_u; @@ -788,8 +834,6 @@ int bch2_inode_rm(struct bch_fs *c, subvol_inum inum) u32 snapshot; int ret; - bch2_trans_init(&trans, c, 0, 1024); - /* * If this was a directory, there shouldn't be any real dirents left - * but there could be whiteouts (from hash collisions) that we should @@ -798,53 +842,61 @@ int bch2_inode_rm(struct bch_fs *c, subvol_inum inum) * XXX: the dirent could ideally would delete whiteouts when they're no * longer needed */ - ret = bch2_inode_delete_keys(&trans, inum, BTREE_ID_extents) ?: - bch2_inode_delete_keys(&trans, inum, BTREE_ID_xattrs) ?: - bch2_inode_delete_keys(&trans, inum, BTREE_ID_dirents); + ret = bch2_inode_delete_keys(trans, inum, BTREE_ID_extents) ?: + bch2_inode_delete_keys(trans, inum, BTREE_ID_xattrs) ?: + bch2_inode_delete_keys(trans, inum, BTREE_ID_dirents); if (ret) goto err; retry: - bch2_trans_begin(&trans); + bch2_trans_begin(trans); - ret = bch2_subvolume_get_snapshot(&trans, inum.subvol, &snapshot); + ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot); if (ret) goto err; - bch2_trans_iter_init(&trans, &iter, BTREE_ID_inodes, - SPOS(0, inum.inum, snapshot), - BTREE_ITER_INTENT|BTREE_ITER_CACHED); - k = bch2_btree_iter_peek_slot(&iter); - + k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes, + SPOS(0, inum.inum, snapshot), + BTREE_ITER_INTENT|BTREE_ITER_CACHED); ret = bkey_err(k); if (ret) goto err; if (!bkey_is_inode(k.k)) { - bch2_fs_inconsistent(trans.c, - "inode %llu not found when deleting", - inum.inum); + bch2_fs_inconsistent(c, + "inode %llu:%u not found when deleting", + inum.inum, snapshot); ret = -EIO; goto err; } bch2_inode_unpack(k, &inode_u); - /* Subvolume root? */ - BUG_ON(inode_u.bi_subvol); - bkey_inode_generation_init(&delete.k_i); delete.k.p = iter.pos; delete.v.bi_generation = cpu_to_le32(inode_u.bi_generation + 1); - ret = bch2_trans_update(&trans, &iter, &delete.k_i, 0) ?: - bch2_trans_commit(&trans, NULL, NULL, + ret = bch2_trans_update(trans, &iter, &delete.k_i, 0) ?: + bch2_trans_commit(trans, NULL, NULL, BTREE_INSERT_NOFAIL); err: - bch2_trans_iter_exit(&trans, &iter); + bch2_trans_iter_exit(trans, &iter); if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) goto retry; - bch2_trans_exit(&trans); + bch2_trans_put(trans); + return ret; +} + +int bch2_inode_find_by_inum_nowarn_trans(struct btree_trans *trans, + subvol_inum inum, + struct bch_inode_unpacked *inode) +{ + struct btree_iter iter; + int ret; + + ret = bch2_inode_peek_nowarn(trans, &iter, inode, inum, 0); + if (!ret) + bch2_trans_iter_exit(trans, &iter); return ret; } @@ -865,7 +917,7 @@ int bch2_inode_find_by_inum(struct bch_fs *c, subvol_inum inum, struct bch_inode_unpacked *inode) { return bch2_trans_do(c, NULL, NULL, 0, - bch2_inode_find_by_inum_trans(&trans, inum, inode)); + bch2_inode_find_by_inum_trans(trans, inum, inode)); } int bch2_inode_nlink_inc(struct bch_inode_unpacked *bi) @@ -922,3 +974,156 @@ void bch2_inode_opts_get(struct bch_io_opts *opts, struct bch_fs *c, if (opts->nocow) opts->compression = opts->background_compression = opts->data_checksum = opts->erasure_code = 0; } + +int bch2_inode_rm_snapshot(struct btree_trans *trans, u64 inum, u32 snapshot) +{ + struct bch_fs *c = trans->c; + struct btree_iter iter = { NULL }; + struct bkey_i_inode_generation delete; + struct bch_inode_unpacked inode_u; + struct bkey_s_c k; + int ret; + + do { + ret = bch2_btree_delete_range_trans(trans, BTREE_ID_extents, + SPOS(inum, 0, snapshot), + SPOS(inum, U64_MAX, snapshot), + 0, NULL) ?: + bch2_btree_delete_range_trans(trans, BTREE_ID_dirents, + SPOS(inum, 0, snapshot), + SPOS(inum, U64_MAX, snapshot), + 0, NULL) ?: + bch2_btree_delete_range_trans(trans, BTREE_ID_xattrs, + SPOS(inum, 0, snapshot), + SPOS(inum, U64_MAX, snapshot), + 0, NULL); + } while (ret == -BCH_ERR_transaction_restart_nested); + if (ret) + goto err; +retry: + bch2_trans_begin(trans); + + k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes, + SPOS(0, inum, snapshot), BTREE_ITER_INTENT); + ret = bkey_err(k); + if (ret) + goto err; + + if (!bkey_is_inode(k.k)) { + bch2_fs_inconsistent(c, + "inode %llu:%u not found when deleting", + inum, snapshot); + ret = -EIO; + goto err; + } + + bch2_inode_unpack(k, &inode_u); + + /* Subvolume root? */ + if (inode_u.bi_subvol) + bch_warn(c, "deleting inode %llu marked as unlinked, but also a subvolume root!?", inode_u.bi_inum); + + bkey_inode_generation_init(&delete.k_i); + delete.k.p = iter.pos; + delete.v.bi_generation = cpu_to_le32(inode_u.bi_generation + 1); + + ret = bch2_trans_update(trans, &iter, &delete.k_i, 0) ?: + bch2_trans_commit(trans, NULL, NULL, + BTREE_INSERT_NOFAIL); +err: + bch2_trans_iter_exit(trans, &iter); + if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) + goto retry; + + return ret ?: -BCH_ERR_transaction_restart_nested; +} + +static int may_delete_deleted_inode(struct btree_trans *trans, struct bpos pos) +{ + struct bch_fs *c = trans->c; + struct btree_iter iter; + struct bkey_s_c k; + struct bch_inode_unpacked inode; + int ret; + + if (bch2_snapshot_is_internal_node(c, pos.snapshot)) + return 0; + + if (!fsck_err_on(c->sb.clean, c, + "filesystem marked as clean but have deleted inode %llu:%u", + pos.offset, pos.snapshot)) + return 0; + + k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes, pos, BTREE_ITER_CACHED); + ret = bkey_err(k); + if (ret) + return ret; + + ret = bkey_is_inode(k.k) ? 0 : -BCH_ERR_ENOENT_inode; + if (fsck_err_on(!bkey_is_inode(k.k), c, + "nonexistent inode %llu:%u in deleted_inodes btree", + pos.offset, pos.snapshot)) + goto delete; + + ret = bch2_inode_unpack(k, &inode); + if (ret) + goto err; + + if (fsck_err_on(S_ISDIR(inode.bi_mode), c, + "directory %llu:%u in deleted_inodes btree", + pos.offset, pos.snapshot)) + goto delete; + + if (fsck_err_on(!(inode.bi_flags & BCH_INODE_UNLINKED), c, + "non-deleted inode %llu:%u in deleted_inodes btree", + pos.offset, pos.snapshot)) + goto delete; + + return 1; +err: +fsck_err: + return ret; +delete: + return bch2_btree_bit_mod(trans, BTREE_ID_deleted_inodes, pos, false); +} + +int bch2_delete_dead_inodes(struct bch_fs *c) +{ + struct btree_trans *trans = bch2_trans_get(c); + struct btree_iter iter; + struct bkey_s_c k; + int ret; + + ret = bch2_btree_write_buffer_flush_sync(trans); + if (ret) + goto err; + + /* + * Weird transaction restart handling here because on successful delete, + * bch2_inode_rm_snapshot() will return a nested transaction restart, + * but we can't retry because the btree write buffer won't have been + * flushed and we'd spin: + */ + for_each_btree_key(trans, iter, BTREE_ID_deleted_inodes, POS_MIN, + BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k, ret) { + ret = lockrestart_do(trans, may_delete_deleted_inode(trans, k.k->p)); + if (ret < 0) + break; + + if (ret) { + if (!test_bit(BCH_FS_RW, &c->flags)) { + bch2_trans_unlock(trans); + bch2_fs_lazy_rw(c); + } + + ret = bch2_inode_rm_snapshot(trans, k.k->p.offset, k.k->p.snapshot); + if (ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart)) + break; + } + } + bch2_trans_iter_exit(trans, &iter); +err: + bch2_trans_put(trans); + + return ret; +}