X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fdirent.c;h=d37bd07afbfe4088ebb9b92feb56ff86127ae1ab;hb=HEAD;hp=de528c3afe8c56faea357392e3ea0a7eaad84c49;hpb=f1f88825c371f84edb85a156de5e1962503d23b2;p=bcachefs-tools-debian diff --git a/libbcachefs/dirent.c b/libbcachefs/dirent.c index de528c3..d37bd07 100644 --- a/libbcachefs/dirent.c +++ b/libbcachefs/dirent.c @@ -1,6 +1,7 @@ // SPDX-License-Identifier: GPL-2.0 #include "bcachefs.h" +#include "bkey_buf.h" #include "bkey_methods.h" #include "btree_update.h" #include "extents.h" @@ -12,12 +13,25 @@ #include -unsigned bch2_dirent_name_bytes(struct bkey_s_c_dirent d) +static unsigned bch2_dirent_name_bytes(struct bkey_s_c_dirent d) { - unsigned len = bkey_val_bytes(d.k) - - offsetof(struct bch_dirent, d_name); + unsigned bkey_u64s = bkey_val_u64s(d.k); + unsigned bkey_bytes = bkey_u64s * sizeof(u64); + u64 last_u64 = ((u64*)d.v)[bkey_u64s - 1]; +#if CPU_BIG_ENDIAN + unsigned trailing_nuls = last_u64 ? __builtin_ctzll(last_u64) / 8 : 64 / 8; +#else + unsigned trailing_nuls = last_u64 ? __builtin_clzll(last_u64) / 8 : 64 / 8; +#endif + + return bkey_bytes - + offsetof(struct bch_dirent, d_name) - + trailing_nuls; +} - return strnlen(d.v->d_name, len); +struct qstr bch2_dirent_get_name(struct bkey_s_c_dirent d) +{ + return (struct qstr) QSTR_INIT(d.v->d_name, bch2_dirent_name_bytes(d)); } static u64 bch2_dirent_hash(const struct bch_hash_info *info, @@ -40,7 +54,7 @@ static u64 dirent_hash_key(const struct bch_hash_info *info, const void *key) static u64 dirent_hash_bkey(const struct bch_hash_info *info, struct bkey_s_c k) { struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k); - struct qstr name = QSTR_INIT(d.v->d_name, bch2_dirent_name_bytes(d)); + struct qstr name = bch2_dirent_get_name(d); return bch2_dirent_hash(info, &name); } @@ -48,20 +62,20 @@ static u64 dirent_hash_bkey(const struct bch_hash_info *info, struct bkey_s_c k) static bool dirent_cmp_key(struct bkey_s_c _l, const void *_r) { struct bkey_s_c_dirent l = bkey_s_c_to_dirent(_l); - int len = bch2_dirent_name_bytes(l); - const struct qstr *r = _r; + const struct qstr l_name = bch2_dirent_get_name(l); + const struct qstr *r_name = _r; - return len - r->len ?: memcmp(l.v->d_name, r->name, len); + return !qstr_eq(l_name, *r_name); } static bool dirent_cmp_bkey(struct bkey_s_c _l, struct bkey_s_c _r) { struct bkey_s_c_dirent l = bkey_s_c_to_dirent(_l); struct bkey_s_c_dirent r = bkey_s_c_to_dirent(_r); - int l_len = bch2_dirent_name_bytes(l); - int r_len = bch2_dirent_name_bytes(r); + const struct qstr l_name = bch2_dirent_get_name(l); + const struct qstr r_name = bch2_dirent_get_name(r); - return l_len - r_len ?: memcmp(l.v->d_name, r.v->d_name, l_len); + return !qstr_eq(l_name, r_name); } static bool dirent_is_visible(subvol_inum inum, struct bkey_s_c k) @@ -83,72 +97,68 @@ const struct bch_hash_desc bch2_dirent_hash_desc = { .is_visible = dirent_is_visible, }; -int bch2_dirent_invalid(const struct bch_fs *c, struct bkey_s_c k, - int rw, struct printbuf *err) +int bch2_dirent_invalid(struct bch_fs *c, struct bkey_s_c k, + enum bkey_invalid_flags flags, + struct printbuf *err) { struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k); - unsigned len; - - if (bkey_val_bytes(k.k) < sizeof(struct bch_dirent)) { - prt_printf(err, "incorrect value size (%zu < %zu)", - bkey_val_bytes(k.k), sizeof(*d.v)); - return -BCH_ERR_invalid_bkey; - } - - len = bch2_dirent_name_bytes(d); - if (!len) { - prt_printf(err, "empty name"); - return -BCH_ERR_invalid_bkey; - } - - if (bkey_val_u64s(k.k) > dirent_val_u64s(len)) { - prt_printf(err, "value too big (%zu > %u)", - bkey_val_u64s(k.k), dirent_val_u64s(len)); - return -BCH_ERR_invalid_bkey; - } - - if (len > BCH_NAME_MAX) { - prt_printf(err, "dirent name too big (%u > %u)", - len, BCH_NAME_MAX); - return -BCH_ERR_invalid_bkey; - } - - if (len == 1 && !memcmp(d.v->d_name, ".", 1)) { - prt_printf(err, "invalid name"); - return -BCH_ERR_invalid_bkey; - } - - if (len == 2 && !memcmp(d.v->d_name, "..", 2)) { - prt_printf(err, "invalid name"); - return -BCH_ERR_invalid_bkey; - } + struct qstr d_name = bch2_dirent_get_name(d); + int ret = 0; - if (memchr(d.v->d_name, '/', len)) { - prt_printf(err, "invalid name"); - return -BCH_ERR_invalid_bkey; - } + bkey_fsck_err_on(!d_name.len, c, err, + dirent_empty_name, + "empty name"); - if (d.v->d_type != DT_SUBVOL && - le64_to_cpu(d.v->d_inum) == d.k->p.inode) { - prt_printf(err, "dirent points to own directory"); - return -BCH_ERR_invalid_bkey; - } + bkey_fsck_err_on(bkey_val_u64s(k.k) > dirent_val_u64s(d_name.len), c, err, + dirent_val_too_big, + "value too big (%zu > %u)", + bkey_val_u64s(k.k), dirent_val_u64s(d_name.len)); - return 0; + /* + * Check new keys don't exceed the max length + * (older keys may be larger.) + */ + bkey_fsck_err_on((flags & BKEY_INVALID_COMMIT) && d_name.len > BCH_NAME_MAX, c, err, + dirent_name_too_long, + "dirent name too big (%u > %u)", + d_name.len, BCH_NAME_MAX); + + bkey_fsck_err_on(d_name.len != strnlen(d_name.name, d_name.len), c, err, + dirent_name_embedded_nul, + "dirent has stray data after name's NUL"); + + bkey_fsck_err_on((d_name.len == 1 && !memcmp(d_name.name, ".", 1)) || + (d_name.len == 2 && !memcmp(d_name.name, "..", 2)), c, err, + dirent_name_dot_or_dotdot, + "invalid name"); + + bkey_fsck_err_on(memchr(d_name.name, '/', d_name.len), c, err, + dirent_name_has_slash, + "name with /"); + + bkey_fsck_err_on(d.v->d_type != DT_SUBVOL && + le64_to_cpu(d.v->d_inum) == d.k->p.inode, c, err, + dirent_to_itself, + "dirent points to own directory"); +fsck_err: + return ret; } -void bch2_dirent_to_text(struct printbuf *out, struct bch_fs *c, - struct bkey_s_c k) +void bch2_dirent_to_text(struct printbuf *out, struct bch_fs *c, struct bkey_s_c k) { struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k); + struct qstr d_name = bch2_dirent_get_name(d); + + prt_printf(out, "%.*s -> ", d_name.len, d_name.name); - prt_printf(out, "%.*s -> %llu type %s", - bch2_dirent_name_bytes(d), - d.v->d_name, - d.v->d_type != DT_SUBVOL - ? le64_to_cpu(d.v->d_inum) - : le32_to_cpu(d.v->d_child_subvol), - bch2_d_type_str(d.v->d_type)); + if (d.v->d_type != DT_SUBVOL) + prt_printf(out, "%llu", le64_to_cpu(d.v->d_inum)); + else + prt_printf(out, "%u -> %u", + le32_to_cpu(d.v->d_parent_subvol), + le32_to_cpu(d.v->d_child_subvol)); + + prt_printf(out, " type %s", bch2_d_type_str(d.v->d_type)); } static struct bkey_i_dirent *dirent_create_key(struct btree_trans *trans, @@ -190,10 +200,39 @@ static struct bkey_i_dirent *dirent_create_key(struct btree_trans *trans, return dirent; } +int bch2_dirent_create_snapshot(struct btree_trans *trans, + u32 dir_subvol, u64 dir, u32 snapshot, + const struct bch_hash_info *hash_info, + u8 type, const struct qstr *name, u64 dst_inum, + u64 *dir_offset, + bch_str_hash_flags_t str_hash_flags) +{ + subvol_inum dir_inum = { .subvol = dir_subvol, .inum = dir }; + struct bkey_i_dirent *dirent; + int ret; + + dirent = dirent_create_key(trans, dir_inum, type, name, dst_inum); + ret = PTR_ERR_OR_ZERO(dirent); + if (ret) + return ret; + + dirent->k.p.inode = dir; + dirent->k.p.snapshot = snapshot; + + ret = bch2_hash_set_in_snapshot(trans, bch2_dirent_hash_desc, hash_info, + dir_inum, snapshot, + &dirent->k_i, str_hash_flags, + BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE); + *dir_offset = dirent->k.p.offset; + + return ret; +} + int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir, const struct bch_hash_info *hash_info, u8 type, const struct qstr *name, u64 dst_inum, - u64 *dir_offset, int flags) + u64 *dir_offset, + bch_str_hash_flags_t str_hash_flags) { struct bkey_i_dirent *dirent; int ret; @@ -204,7 +243,7 @@ int bch2_dirent_create(struct btree_trans *trans, subvol_inum dir, return ret; ret = bch2_hash_set(trans, bch2_dirent_hash_desc, hash_info, - dir, &dirent->k_i, flags); + dir, &dirent->k_i, str_hash_flags); *dir_offset = dirent->k.p.offset; return ret; @@ -224,7 +263,7 @@ int bch2_dirent_read_target(struct btree_trans *trans, subvol_inum dir, int ret = 0; if (d.v->d_type == DT_SUBVOL && - d.v->d_parent_subvol != dir.subvol) + le32_to_cpu(d.v->d_parent_subvol) != dir.subvol) return 1; if (likely(d.v->d_type != DT_SUBVOL)) { @@ -254,12 +293,10 @@ int bch2_dirent_rename(struct btree_trans *trans, struct bkey_i_dirent *new_src = NULL, *new_dst = NULL; struct bpos dst_pos = POS(dst_dir.inum, bch2_dirent_hash(dst_hash, dst_name)); - unsigned src_type = 0, dst_type = 0, src_update_flags = 0; + unsigned src_update_flags = 0; + bool delete_src, delete_dst; int ret = 0; - if (src_dir.subvol != dst_dir.subvol) - return -EXDEV; - memset(src_inum, 0, sizeof(*src_inum)); memset(dst_inum, 0, sizeof(*dst_inum)); @@ -280,12 +317,6 @@ int bch2_dirent_rename(struct btree_trans *trans, if (ret) goto out; - src_type = bkey_s_c_to_dirent(old_src).v->d_type; - - if (src_type == DT_SUBVOL && mode == BCH_RENAME_EXCHANGE) - return -EOPNOTSUPP; - - /* Lookup dst: */ if (mode == BCH_RENAME) { /* @@ -313,11 +344,6 @@ int bch2_dirent_rename(struct btree_trans *trans, bkey_s_c_to_dirent(old_dst), dst_inum); if (ret) goto out; - - dst_type = bkey_s_c_to_dirent(old_dst).v->d_type; - - if (dst_type == DT_SUBVOL) - return -EOPNOTSUPP; } if (mode != BCH_RENAME_EXCHANGE) @@ -350,8 +376,8 @@ int bch2_dirent_rename(struct btree_trans *trans, bkey_init(&new_src->k); new_src->k.p = src_iter.pos; - if (bkey_cmp(dst_pos, src_iter.pos) <= 0 && - bkey_cmp(src_iter.pos, dst_iter.pos) < 0) { + if (bkey_le(dst_pos, src_iter.pos) && + bkey_lt(src_iter.pos, dst_iter.pos)) { /* * We have a hash collision for the new dst key, * and new_src - the key we're deleting - is between @@ -387,28 +413,55 @@ int bch2_dirent_rename(struct btree_trans *trans, } } + if (new_dst->v.d_type == DT_SUBVOL) + new_dst->v.d_parent_subvol = cpu_to_le32(dst_dir.subvol); + + if ((mode == BCH_RENAME_EXCHANGE) && + new_src->v.d_type == DT_SUBVOL) + new_src->v.d_parent_subvol = cpu_to_le32(src_dir.subvol); + ret = bch2_trans_update(trans, &dst_iter, &new_dst->k_i, 0); if (ret) goto out; out_set_src: - /* - * If we're deleting a subvolume, we need to really delete the dirent, - * not just emit a whiteout in the current snapshot: + * If we're deleting a subvolume we need to really delete the dirent, + * not just emit a whiteout in the current snapshot - there can only be + * single dirent that points to a given subvolume. + * + * IOW, we don't maintain multiple versions in different snapshots of + * dirents that point to subvolumes - dirents that point to subvolumes + * are only visible in one particular subvolume so it's not necessary, + * and it would be particularly confusing for fsck to have to deal with. */ - if (src_type == DT_SUBVOL) { - bch2_btree_iter_set_snapshot(&src_iter, old_src.k->p.snapshot); - ret = bch2_btree_iter_traverse(&src_iter); + delete_src = bkey_s_c_to_dirent(old_src).v->d_type == DT_SUBVOL && + new_src->k.p.snapshot != old_src.k->p.snapshot; + + delete_dst = old_dst.k && + bkey_s_c_to_dirent(old_dst).v->d_type == DT_SUBVOL && + new_dst->k.p.snapshot != old_dst.k->p.snapshot; + + if (!delete_src || !bkey_deleted(&new_src->k)) { + ret = bch2_trans_update(trans, &src_iter, &new_src->k_i, src_update_flags); if (ret) goto out; + } - new_src->k.p = src_iter.pos; - src_update_flags |= BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE; + if (delete_src) { + bch2_btree_iter_set_snapshot(&src_iter, old_src.k->p.snapshot); + ret = bch2_btree_iter_traverse(&src_iter) ?: + bch2_btree_delete_at(trans, &src_iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE); + if (ret) + goto out; } - ret = bch2_trans_update(trans, &src_iter, &new_src->k_i, src_update_flags); - if (ret) - goto out; + if (delete_dst) { + bch2_btree_iter_set_snapshot(&dst_iter, old_dst.k->p.snapshot); + ret = bch2_btree_iter_traverse(&dst_iter) ?: + bch2_btree_delete_at(trans, &dst_iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE); + if (ret) + goto out; + } if (mode == BCH_RENAME_EXCHANGE) *src_offset = new_src->k.p.offset; @@ -419,41 +472,29 @@ out: return ret; } -int __bch2_dirent_lookup_trans(struct btree_trans *trans, - struct btree_iter *iter, - subvol_inum dir, - const struct bch_hash_info *hash_info, - const struct qstr *name, subvol_inum *inum, - unsigned flags) +int bch2_dirent_lookup_trans(struct btree_trans *trans, + struct btree_iter *iter, + subvol_inum dir, + const struct bch_hash_info *hash_info, + const struct qstr *name, subvol_inum *inum, + unsigned flags) { - struct bkey_s_c k; - struct bkey_s_c_dirent d; - u32 snapshot; - int ret; - - ret = bch2_subvolume_get_snapshot(trans, dir.subvol, &snapshot); + int ret = bch2_hash_lookup(trans, iter, bch2_dirent_hash_desc, + hash_info, dir, name, flags); if (ret) return ret; - ret = bch2_hash_lookup(trans, iter, bch2_dirent_hash_desc, - hash_info, dir, name, flags); - if (ret) - return ret; - - k = bch2_btree_iter_peek_slot(iter); + struct bkey_s_c k = bch2_btree_iter_peek_slot(iter); ret = bkey_err(k); if (ret) goto err; - d = bkey_s_c_to_dirent(k); - - ret = bch2_dirent_read_target(trans, dir, d, inum); + ret = bch2_dirent_read_target(trans, dir, bkey_s_c_to_dirent(k), inum); if (ret > 0) ret = -ENOENT; err: if (ret) bch2_trans_iter_exit(trans, iter); - return ret; } @@ -461,40 +502,30 @@ u64 bch2_dirent_lookup(struct bch_fs *c, subvol_inum dir, const struct bch_hash_info *hash_info, const struct qstr *name, subvol_inum *inum) { - struct btree_trans trans; - struct btree_iter iter; - int ret; + struct btree_trans *trans = bch2_trans_get(c); + struct btree_iter iter = { NULL }; - bch2_trans_init(&trans, c, 0, 0); -retry: - bch2_trans_begin(&trans); - - ret = __bch2_dirent_lookup_trans(&trans, &iter, dir, hash_info, - name, inum, 0); - if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) - goto retry; - if (!ret) - bch2_trans_iter_exit(&trans, &iter); - bch2_trans_exit(&trans); + int ret = lockrestart_do(trans, + bch2_dirent_lookup_trans(trans, &iter, dir, hash_info, name, inum, 0)); + bch2_trans_iter_exit(trans, &iter); + bch2_trans_put(trans); return ret; } -int bch2_empty_dir_trans(struct btree_trans *trans, subvol_inum dir) +int bch2_empty_dir_snapshot(struct btree_trans *trans, u64 dir, u32 subvol, u32 snapshot) { struct btree_iter iter; struct bkey_s_c k; - u32 snapshot; int ret; - ret = bch2_subvolume_get_snapshot(trans, dir.subvol, &snapshot); - if (ret) - return ret; - for_each_btree_key_upto_norestart(trans, iter, BTREE_ID_dirents, - SPOS(dir.inum, 0, snapshot), - POS(dir.inum, U64_MAX), 0, k, ret) + SPOS(dir, 0, snapshot), + POS(dir, U64_MAX), 0, k, ret) if (k.k->type == KEY_TYPE_dirent) { - ret = -ENOTEMPTY; + struct bkey_s_c_dirent d = bkey_s_c_to_dirent(k); + if (d.v->d_type == DT_SUBVOL && le32_to_cpu(d.v->d_parent_subvol) != subvol) + continue; + ret = -BCH_ERR_ENOTEMPTY_dir_not_empty; break; } bch2_trans_iter_exit(trans, &iter); @@ -502,25 +533,35 @@ int bch2_empty_dir_trans(struct btree_trans *trans, subvol_inum dir) return ret; } +int bch2_empty_dir_trans(struct btree_trans *trans, subvol_inum dir) +{ + u32 snapshot; + + return bch2_subvolume_get_snapshot(trans, dir.subvol, &snapshot) ?: + bch2_empty_dir_snapshot(trans, dir.inum, dir.subvol, snapshot); +} + int bch2_readdir(struct bch_fs *c, subvol_inum inum, struct dir_context *ctx) { - struct btree_trans trans; + struct btree_trans *trans = bch2_trans_get(c); struct btree_iter iter; struct bkey_s_c k; struct bkey_s_c_dirent dirent; subvol_inum target; u32 snapshot; + struct bkey_buf sk; + struct qstr name; int ret; - bch2_trans_init(&trans, c, 0, 0); + bch2_bkey_buf_init(&sk); 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; - for_each_btree_key_upto_norestart(&trans, iter, BTREE_ID_dirents, + for_each_btree_key_upto_norestart(trans, iter, BTREE_ID_dirents, SPOS(inum.inum, ctx->pos, snapshot), POS(inum.inum, U64_MAX), 0, k, ret) { if (k.k->type != KEY_TYPE_dirent) @@ -528,19 +569,22 @@ retry: dirent = bkey_s_c_to_dirent(k); - ret = bch2_dirent_read_target(&trans, inum, dirent, &target); + ret = bch2_dirent_read_target(trans, inum, dirent, &target); if (ret < 0) break; if (ret) continue; - /* - * XXX: dir_emit() can fault and block, while we're holding - * locks - */ + /* dir_emit() can fault and block: */ + bch2_bkey_buf_reassemble(&sk, c, k); + dirent = bkey_i_to_s_c_dirent(sk.k); + bch2_trans_unlock(trans); + + name = bch2_dirent_get_name(dirent); + ctx->pos = dirent.k->p.offset; - if (!dir_emit(ctx, dirent.v->d_name, - bch2_dirent_name_bytes(dirent), + if (!dir_emit(ctx, name.name, + name.len, target.inum, vfs_d_type(dirent.v->d_type))) break; @@ -550,16 +594,17 @@ retry: * read_target looks up subvolumes, we can overflow paths if the * directory has many subvolumes in it */ - ret = btree_trans_too_many_iters(&trans); + ret = btree_trans_too_many_iters(trans); if (ret) break; } - bch2_trans_iter_exit(&trans, &iter); + bch2_trans_iter_exit(trans, &iter); err: if (bch2_err_matches(ret, BCH_ERR_transaction_restart)) goto retry; - bch2_trans_exit(&trans); + bch2_trans_put(trans); + bch2_bkey_buf_exit(&sk, c); return ret; }