diff options
Diffstat (limited to 'fs/bcachefs')
65 files changed, 1197 insertions, 941 deletions
diff --git a/fs/bcachefs/Kconfig b/fs/bcachefs/Kconfig index bf1c94e51dd0..07709b0d7688 100644 --- a/fs/bcachefs/Kconfig +++ b/fs/bcachefs/Kconfig @@ -4,7 +4,7 @@ config BCACHEFS_FS depends on BLOCK select EXPORTFS select CLOSURES - select LIBCRC32C + select CRC32 select CRC64 select FS_POSIX_ACL select LZ4_COMPRESS @@ -15,10 +15,9 @@ config BCACHEFS_FS select ZLIB_INFLATE select ZSTD_COMPRESS select ZSTD_DECOMPRESS - select CRYPTO select CRYPTO_LIB_SHA256 - select CRYPTO_CHACHA20 - select CRYPTO_POLY1305 + select CRYPTO_LIB_CHACHA + select CRYPTO_LIB_POLY1305 select KEYS select RAID6_PQ select XOR_BLOCKS diff --git a/fs/bcachefs/alloc_foreground.c b/fs/bcachefs/alloc_foreground.c index 7c930ef77380..7ec022e9361a 100644 --- a/fs/bcachefs/alloc_foreground.c +++ b/fs/bcachefs/alloc_foreground.c @@ -1422,8 +1422,30 @@ alloc_done: wp->sectors_free = UINT_MAX; - open_bucket_for_each(c, &wp->ptrs, ob, i) + open_bucket_for_each(c, &wp->ptrs, ob, i) { + /* + * Ensure proper write alignment - either due to misaligned + * bucket sizes (from buggy bcachefs-tools), or writes that mix + * logical/physical alignment: + */ + struct bch_dev *ca = ob_dev(c, ob); + u64 offset = bucket_to_sector(ca, ob->bucket) + + ca->mi.bucket_size - + ob->sectors_free; + unsigned align = round_up(offset, block_sectors(c)) - offset; + + ob->sectors_free = max_t(int, 0, ob->sectors_free - align); + wp->sectors_free = min(wp->sectors_free, ob->sectors_free); + } + + wp->sectors_free = rounddown(wp->sectors_free, block_sectors(c)); + + /* Did alignment use up space in an open_bucket? */ + if (unlikely(!wp->sectors_free)) { + bch2_alloc_sectors_done(c, wp); + goto retry; + } BUG_ON(!wp->sectors_free || wp->sectors_free == UINT_MAX); diff --git a/fs/bcachefs/alloc_foreground.h b/fs/bcachefs/alloc_foreground.h index 69ec6a012898..4c1e33cf57c0 100644 --- a/fs/bcachefs/alloc_foreground.h +++ b/fs/bcachefs/alloc_foreground.h @@ -110,7 +110,9 @@ static inline void bch2_alloc_sectors_done_inlined(struct bch_fs *c, struct writ unsigned i; open_bucket_for_each(c, &wp->ptrs, ob, i) - ob_push(c, !ob->sectors_free ? &ptrs : &keep, ob); + ob_push(c, ob->sectors_free < block_sectors(c) + ? &ptrs + : &keep, ob); wp->ptrs = keep; mutex_unlock(&wp->lock); diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h index 5d9f208a1bb7..75f7408da173 100644 --- a/fs/bcachefs/bcachefs.h +++ b/fs/bcachefs/bcachefs.h @@ -788,6 +788,8 @@ struct bch_fs { unsigned long errors_silent[BITS_TO_LONGS(BCH_FSCK_ERR_MAX)]; u64 btrees_lost_data; } sb; + DARRAY(enum bcachefs_metadata_version) + incompat_versions_requested; #ifdef CONFIG_UNICODE struct unicode_map *cf_encoding; @@ -981,8 +983,8 @@ struct bch_fs { mempool_t compress_workspace[BCH_COMPRESSION_OPT_NR]; size_t zstd_workspace_size; - struct crypto_sync_skcipher *chacha20; - struct crypto_shash *poly1305; + struct bch_key chacha20_key; + bool chacha20_key_set; atomic64_t key_version; diff --git a/fs/bcachefs/bcachefs_format.h b/fs/bcachefs/bcachefs_format.h index a3db328dee31..d6e4a496f02b 100644 --- a/fs/bcachefs/bcachefs_format.h +++ b/fs/bcachefs/bcachefs_format.h @@ -366,6 +366,10 @@ static inline void bkey_init(struct bkey *k) #define __BKEY_PADDED(key, pad) \ struct bkey_i key; __u64 key ## _pad[pad] +enum bch_bkey_type_flags { + BKEY_TYPE_strict_btree_checks = BIT(0), +}; + /* * - DELETED keys are used internally to mark keys that should be ignored but * override keys in composition order. Their version number is ignored. @@ -383,46 +387,46 @@ static inline void bkey_init(struct bkey *k) * * - WHITEOUT: for hash table btrees */ -#define BCH_BKEY_TYPES() \ - x(deleted, 0) \ - x(whiteout, 1) \ - x(error, 2) \ - x(cookie, 3) \ - x(hash_whiteout, 4) \ - x(btree_ptr, 5) \ - x(extent, 6) \ - x(reservation, 7) \ - x(inode, 8) \ - x(inode_generation, 9) \ - x(dirent, 10) \ - x(xattr, 11) \ - x(alloc, 12) \ - x(quota, 13) \ - x(stripe, 14) \ - x(reflink_p, 15) \ - x(reflink_v, 16) \ - x(inline_data, 17) \ - x(btree_ptr_v2, 18) \ - x(indirect_inline_data, 19) \ - x(alloc_v2, 20) \ - x(subvolume, 21) \ - x(snapshot, 22) \ - x(inode_v2, 23) \ - x(alloc_v3, 24) \ - x(set, 25) \ - x(lru, 26) \ - x(alloc_v4, 27) \ - x(backpointer, 28) \ - x(inode_v3, 29) \ - x(bucket_gens, 30) \ - x(snapshot_tree, 31) \ - x(logged_op_truncate, 32) \ - x(logged_op_finsert, 33) \ - x(accounting, 34) \ - x(inode_alloc_cursor, 35) +#define BCH_BKEY_TYPES() \ + x(deleted, 0, 0) \ + x(whiteout, 1, 0) \ + x(error, 2, 0) \ + x(cookie, 3, 0) \ + x(hash_whiteout, 4, BKEY_TYPE_strict_btree_checks) \ + x(btree_ptr, 5, BKEY_TYPE_strict_btree_checks) \ + x(extent, 6, BKEY_TYPE_strict_btree_checks) \ + x(reservation, 7, BKEY_TYPE_strict_btree_checks) \ + x(inode, 8, BKEY_TYPE_strict_btree_checks) \ + x(inode_generation, 9, BKEY_TYPE_strict_btree_checks) \ + x(dirent, 10, BKEY_TYPE_strict_btree_checks) \ + x(xattr, 11, BKEY_TYPE_strict_btree_checks) \ + x(alloc, 12, BKEY_TYPE_strict_btree_checks) \ + x(quota, 13, BKEY_TYPE_strict_btree_checks) \ + x(stripe, 14, BKEY_TYPE_strict_btree_checks) \ + x(reflink_p, 15, BKEY_TYPE_strict_btree_checks) \ + x(reflink_v, 16, BKEY_TYPE_strict_btree_checks) \ + x(inline_data, 17, BKEY_TYPE_strict_btree_checks) \ + x(btree_ptr_v2, 18, BKEY_TYPE_strict_btree_checks) \ + x(indirect_inline_data, 19, BKEY_TYPE_strict_btree_checks) \ + x(alloc_v2, 20, BKEY_TYPE_strict_btree_checks) \ + x(subvolume, 21, BKEY_TYPE_strict_btree_checks) \ + x(snapshot, 22, BKEY_TYPE_strict_btree_checks) \ + x(inode_v2, 23, BKEY_TYPE_strict_btree_checks) \ + x(alloc_v3, 24, BKEY_TYPE_strict_btree_checks) \ + x(set, 25, 0) \ + x(lru, 26, BKEY_TYPE_strict_btree_checks) \ + x(alloc_v4, 27, BKEY_TYPE_strict_btree_checks) \ + x(backpointer, 28, BKEY_TYPE_strict_btree_checks) \ + x(inode_v3, 29, BKEY_TYPE_strict_btree_checks) \ + x(bucket_gens, 30, BKEY_TYPE_strict_btree_checks) \ + x(snapshot_tree, 31, BKEY_TYPE_strict_btree_checks) \ + x(logged_op_truncate, 32, BKEY_TYPE_strict_btree_checks) \ + x(logged_op_finsert, 33, BKEY_TYPE_strict_btree_checks) \ + x(accounting, 34, BKEY_TYPE_strict_btree_checks) \ + x(inode_alloc_cursor, 35, BKEY_TYPE_strict_btree_checks) enum bch_bkey_type { -#define x(name, nr) KEY_TYPE_##name = nr, +#define x(name, nr, ...) KEY_TYPE_##name = nr, BCH_BKEY_TYPES() #undef x KEY_TYPE_MAX, @@ -863,6 +867,7 @@ LE64_BITMASK(BCH_SB_VERSION_INCOMPAT_ALLOWED, LE64_BITMASK(BCH_SB_SHARD_INUMS_NBITS, struct bch_sb, flags[6], 0, 4); LE64_BITMASK(BCH_SB_WRITE_ERROR_TIMEOUT,struct bch_sb, flags[6], 4, 14); LE64_BITMASK(BCH_SB_CSUM_ERR_RETRY_NR, struct bch_sb, flags[6], 14, 20); +LE64_BITMASK(BCH_SB_CASEFOLD, struct bch_sb, flags[6], 22, 23); static inline __u64 BCH_SB_COMPRESSION_TYPE(const struct bch_sb *sb) { diff --git a/fs/bcachefs/bkey_methods.c b/fs/bcachefs/bkey_methods.c index 15c93576b5c2..00d05ccfaf73 100644 --- a/fs/bcachefs/bkey_methods.c +++ b/fs/bcachefs/bkey_methods.c @@ -21,7 +21,7 @@ #include "xattr.h" const char * const bch2_bkey_types[] = { -#define x(name, nr) #name, +#define x(name, nr, ...) #name, BCH_BKEY_TYPES() #undef x NULL @@ -115,7 +115,7 @@ static bool key_type_set_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_ }) const struct bkey_ops bch2_bkey_ops[] = { -#define x(name, nr) [KEY_TYPE_##name] = bch2_bkey_ops_##name, +#define x(name, nr, ...) [KEY_TYPE_##name] = bch2_bkey_ops_##name, BCH_BKEY_TYPES() #undef x }; @@ -155,6 +155,12 @@ static u64 bch2_key_types_allowed[] = { #undef x }; +static const enum bch_bkey_type_flags bch2_bkey_type_flags[] = { +#define x(name, nr, flags) [KEY_TYPE_##name] = flags, + BCH_BKEY_TYPES() +#undef x +}; + const char *bch2_btree_node_type_str(enum btree_node_type type) { return type == BKEY_TYPE_btree ? "internal btree node" : bch2_btree_id_str(type - 1); @@ -177,8 +183,18 @@ int __bch2_bkey_validate(struct bch_fs *c, struct bkey_s_c k, if (type >= BKEY_TYPE_NR) return 0; - bkey_fsck_err_on(k.k->type < KEY_TYPE_MAX && - (type == BKEY_TYPE_btree || (from.flags & BCH_VALIDATE_commit)) && + enum bch_bkey_type_flags bkey_flags = k.k->type < KEY_TYPE_MAX + ? bch2_bkey_type_flags[k.k->type] + : 0; + + bool strict_key_type_allowed = + (from.flags & BCH_VALIDATE_commit) || + type == BKEY_TYPE_btree || + (from.btree < BTREE_ID_NR && + (bkey_flags & BKEY_TYPE_strict_btree_checks)); + + bkey_fsck_err_on(strict_key_type_allowed && + k.k->type < KEY_TYPE_MAX && !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), c, bkey_invalid_type_for_btree, "invalid key type for btree %s (%s)", diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c index 7b98ba2dec64..37b69d89341f 100644 --- a/fs/bcachefs/btree_gc.c +++ b/fs/bcachefs/btree_gc.c @@ -47,6 +47,27 @@ #define DROP_PREV_NODE 11 #define DID_FILL_FROM_SCAN 12 +/* + * Returns true if it's a btree we can easily reconstruct, or otherwise won't + * cause data loss if it's missing: + */ +static bool btree_id_important(enum btree_id btree) +{ + if (btree_id_is_alloc(btree)) + return false; + + switch (btree) { + case BTREE_ID_quotas: + case BTREE_ID_snapshot_trees: + case BTREE_ID_logged_ops: + case BTREE_ID_rebalance_work: + case BTREE_ID_subvolume_children: + return false; + default: + return true; + } +} + static const char * const bch2_gc_phase_strs[] = { #define x(n) #n, GC_PHASES() @@ -534,8 +555,10 @@ reconstruct_root: r->error = 0; if (!bch2_btree_has_scanned_nodes(c, i)) { - mustfix_fsck_err(trans, btree_root_unreadable_and_scan_found_nothing, - "no nodes found for btree %s, continue?", buf.buf); + __fsck_err(trans, + FSCK_CAN_FIX|(!btree_id_important(i) ? FSCK_AUTOFIX : 0), + btree_root_unreadable_and_scan_found_nothing, + "no nodes found for btree %s, continue?", buf.buf); bch2_btree_root_alloc_fake_trans(trans, i, 0); } else { bch2_btree_root_alloc_fake_trans(trans, i, 1); diff --git a/fs/bcachefs/btree_io.c b/fs/bcachefs/btree_io.c index 5fd4a58d2ad2..60782f3e5aec 100644 --- a/fs/bcachefs/btree_io.c +++ b/fs/bcachefs/btree_io.c @@ -41,6 +41,7 @@ void bch2_btree_node_io_unlock(struct btree *b) clear_btree_node_write_in_flight_inner(b); clear_btree_node_write_in_flight(b); + smp_mb__after_atomic(); wake_up_bit(&b->flags, BTREE_NODE_write_in_flight); } @@ -1400,6 +1401,7 @@ start: printbuf_exit(&buf); clear_btree_node_read_in_flight(b); + smp_mb__after_atomic(); wake_up_bit(&b->flags, BTREE_NODE_read_in_flight); } @@ -1595,6 +1597,7 @@ fsck_err: printbuf_exit(&buf); clear_btree_node_read_in_flight(b); + smp_mb__after_atomic(); wake_up_bit(&b->flags, BTREE_NODE_read_in_flight); } @@ -1721,6 +1724,7 @@ void bch2_btree_node_read(struct btree_trans *trans, struct btree *b, set_btree_node_read_error(b); bch2_btree_lost_data(c, b->c.btree_id); clear_btree_node_read_in_flight(b); + smp_mb__after_atomic(); wake_up_bit(&b->flags, BTREE_NODE_read_in_flight); printbuf_exit(&buf); return; @@ -2061,8 +2065,10 @@ static void __btree_node_write_done(struct bch_fs *c, struct btree *b, u64 start if (new & (1U << BTREE_NODE_write_in_flight)) __bch2_btree_node_write(c, b, BTREE_WRITE_ALREADY_STARTED|type); - else + else { + smp_mb__after_atomic(); wake_up_bit(&b->flags, BTREE_NODE_write_in_flight); + } } static void btree_node_write_done(struct bch_fs *c, struct btree *b, u64 start_time) @@ -2175,6 +2181,7 @@ static void btree_node_write_endio(struct bio *bio) } clear_btree_node_write_in_flight_inner(b); + smp_mb__after_atomic(); wake_up_bit(&b->flags, BTREE_NODE_write_in_flight_inner); INIT_WORK(&wb->work, btree_node_write_work); queue_work(c->btree_io_complete_wq, &wb->work); diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c index e34e9598ef25..59fa527ac685 100644 --- a/fs/bcachefs/btree_iter.c +++ b/fs/bcachefs/btree_iter.c @@ -2577,7 +2577,10 @@ struct bkey_s_c bch2_btree_iter_peek_prev_min(struct btree_trans *trans, struct struct bpos end) { if ((iter->flags & (BTREE_ITER_is_extents|BTREE_ITER_filter_snapshots)) && - !bkey_eq(iter->pos, POS_MAX)) { + !bkey_eq(iter->pos, POS_MAX) && + !((iter->flags & BTREE_ITER_is_extents) && + iter->pos.offset == U64_MAX)) { + /* * bkey_start_pos(), for extents, is not monotonically * increasing until after filtering for snapshots: @@ -2602,7 +2605,7 @@ struct bkey_s_c bch2_btree_iter_peek_prev_min(struct btree_trans *trans, struct bch2_trans_verify_not_unlocked_or_in_restart(trans); bch2_btree_iter_verify_entry_exit(iter); - EBUG_ON((iter->flags & BTREE_ITER_filter_snapshots) && bpos_eq(end, POS_MIN)); + EBUG_ON((iter->flags & BTREE_ITER_filter_snapshots) && iter->pos.inode != end.inode); int ret = trans_maybe_inject_restart(trans, _RET_IP_); if (unlikely(ret)) { diff --git a/fs/bcachefs/btree_journal_iter.c b/fs/bcachefs/btree_journal_iter.c index d1ad1a7613c9..ade3b5addd75 100644 --- a/fs/bcachefs/btree_journal_iter.c +++ b/fs/bcachefs/btree_journal_iter.c @@ -288,7 +288,7 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id, .size = max_t(size_t, keys->size, 8) * 2, }; - new_keys.data = kvmalloc_array(new_keys.size, sizeof(new_keys.data[0]), GFP_KERNEL); + new_keys.data = bch2_kvmalloc(new_keys.size * sizeof(new_keys.data[0]), GFP_KERNEL); if (!new_keys.data) { bch_err(c, "%s: error allocating new key array (size %zu)", __func__, new_keys.size); @@ -644,8 +644,6 @@ void bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans, */ static int journal_sort_key_cmp(const void *_l, const void *_r) { - cond_resched(); - const struct journal_key *l = _l; const struct journal_key *r = _r; @@ -689,7 +687,8 @@ void bch2_journal_keys_put(struct bch_fs *c) static void __journal_keys_sort(struct journal_keys *keys) { - sort(keys->data, keys->nr, sizeof(keys->data[0]), journal_sort_key_cmp, NULL); + sort_nonatomic(keys->data, keys->nr, sizeof(keys->data[0]), + journal_sort_key_cmp, NULL); cond_resched(); diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c index 8c9fdb7263fe..86acf037590c 100644 --- a/fs/bcachefs/btree_node_scan.c +++ b/fs/bcachefs/btree_node_scan.c @@ -183,7 +183,7 @@ static void try_read_btree_node(struct find_btree_nodes *f, struct bch_dev *ca, return; if (bch2_csum_type_is_encryption(BSET_CSUM_TYPE(&bn->keys))) { - if (!c->chacha20) + if (!c->chacha20_key_set) return; struct nonce nonce = btree_nonce(&bn->keys, 0); @@ -398,7 +398,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c) bch2_print_string_as_lines(KERN_INFO, buf.buf); } - sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_cookie, NULL); + sort_nonatomic(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_cookie, NULL); dst = 0; darray_for_each(f->nodes, i) { @@ -418,7 +418,7 @@ int bch2_scan_for_btree_nodes(struct bch_fs *c) } f->nodes.nr = dst; - sort(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_pos, NULL); + sort_nonatomic(f->nodes.data, f->nodes.nr, sizeof(f->nodes.data[0]), found_btree_node_cmp_pos, NULL); if (0 && c->opts.verbose) { printbuf_reset(&buf); diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c index 55fbeeb8eaaa..00307356d7c8 100644 --- a/fs/bcachefs/btree_update_interior.c +++ b/fs/bcachefs/btree_update_interior.c @@ -1221,7 +1221,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path, ret = bch2_disk_reservation_get(c, &as->disk_res, (nr_nodes[0] + nr_nodes[1]) * btree_sectors(c), - c->opts.metadata_replicas, + READ_ONCE(c->opts.metadata_replicas), disk_res_flags); if (ret) goto err; @@ -1389,7 +1389,7 @@ static void bch2_insert_fixup_btree_ptr(struct btree_update *as, printbuf_exit(&buf); } -static void +static int bch2_btree_insert_keys_interior(struct btree_update *as, struct btree_trans *trans, struct btree_path *path, @@ -1411,7 +1411,8 @@ bch2_btree_insert_keys_interior(struct btree_update *as, insert = bkey_next(insert)) bch2_insert_fixup_btree_ptr(as, trans, path, b, &node_iter, insert); - if (bch2_btree_node_check_topology(trans, b)) { + int ret = bch2_btree_node_check_topology(trans, b); + if (ret) { struct printbuf buf = PRINTBUF; for (struct bkey_i *k = keys->keys; @@ -1421,11 +1422,15 @@ bch2_btree_insert_keys_interior(struct btree_update *as, prt_newline(&buf); } - panic("%s(): check_topology error: inserted keys\n%s", __func__, buf.buf); + bch2_fs_fatal_error(as->c, "%ps -> %s(): check_topology error %s: inserted keys\n%s", + (void *) _RET_IP_, __func__, bch2_err_str(ret), buf.buf); + dump_stack(); + return ret; } memmove_u64s_down(keys->keys, insert, keys->top_p - insert->_data); keys->top_p -= insert->_data - keys->keys_p; + return 0; } static bool key_deleted_in_insert(struct keylist *insert_keys, struct bpos pos) @@ -1559,11 +1564,11 @@ static void __btree_split_node(struct btree_update *as, * nodes that were coalesced, and thus in the middle of a child node post * coalescing: */ -static void btree_split_insert_keys(struct btree_update *as, - struct btree_trans *trans, - btree_path_idx_t path_idx, - struct btree *b, - struct keylist *keys) +static int btree_split_insert_keys(struct btree_update *as, + struct btree_trans *trans, + btree_path_idx_t path_idx, + struct btree *b, + struct keylist *keys) { struct btree_path *path = trans->paths + path_idx; @@ -1573,8 +1578,12 @@ static void btree_split_insert_keys(struct btree_update *as, bch2_btree_node_iter_init(&node_iter, b, &bch2_keylist_front(keys)->k.p); - bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys); + int ret = bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys); + if (ret) + return ret; } + + return 0; } static int btree_split(struct btree_update *as, struct btree_trans *trans, @@ -1607,8 +1616,10 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans, __btree_split_node(as, trans, b, n, keys); if (keys) { - btree_split_insert_keys(as, trans, path, n1, keys); - btree_split_insert_keys(as, trans, path, n2, keys); + ret = btree_split_insert_keys(as, trans, path, n1, keys) ?: + btree_split_insert_keys(as, trans, path, n2, keys); + if (ret) + goto err; BUG_ON(!bch2_keylist_empty(keys)); } @@ -1654,7 +1665,9 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans, n3->sib_u64s[0] = U16_MAX; n3->sib_u64s[1] = U16_MAX; - btree_split_insert_keys(as, trans, path, n3, &as->parent_keys); + ret = btree_split_insert_keys(as, trans, path, n3, &as->parent_keys); + if (ret) + goto err; } } else { trace_and_count(c, btree_node_compact, trans, b); @@ -1662,7 +1675,9 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans, n1 = bch2_btree_node_alloc_replacement(as, trans, b); if (keys) { - btree_split_insert_keys(as, trans, path, n1, keys); + ret = btree_split_insert_keys(as, trans, path, n1, keys); + if (ret) + goto err; BUG_ON(!bch2_keylist_empty(keys)); } @@ -1809,15 +1824,15 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t goto split; } - ret = bch2_btree_node_check_topology(trans, b); + + ret = bch2_btree_node_check_topology(trans, b) ?: + bch2_btree_insert_keys_interior(as, trans, path, b, + path->l[b->c.level].iter, keys); if (ret) { bch2_btree_node_unlock_write(trans, path, b); return ret; } - bch2_btree_insert_keys_interior(as, trans, path, b, - path->l[b->c.level].iter, keys); - trans_for_each_path_with_node(trans, b, linked, i) bch2_btree_node_iter_peek(&linked->l[b->c.level].iter, b); diff --git a/fs/bcachefs/btree_write_buffer.c b/fs/bcachefs/btree_write_buffer.c index adbe576ec77e..0941fb2c026d 100644 --- a/fs/bcachefs/btree_write_buffer.c +++ b/fs/bcachefs/btree_write_buffer.c @@ -428,10 +428,10 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans) */ trace_and_count(c, write_buffer_flush_slowpath, trans, slowpath, wb->flushing.keys.nr); - sort(wb->flushing.keys.data, - wb->flushing.keys.nr, - sizeof(wb->flushing.keys.data[0]), - wb_key_seq_cmp, NULL); + sort_nonatomic(wb->flushing.keys.data, + wb->flushing.keys.nr, + sizeof(wb->flushing.keys.data[0]), + wb_key_seq_cmp, NULL); darray_for_each(wb->flushing.keys, i) { if (!i->journal_seq) diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c index fea61e60a9ee..31fbc2716d8b 100644 --- a/fs/bcachefs/buckets.c +++ b/fs/bcachefs/buckets.c @@ -37,7 +37,8 @@ void bch2_dev_usage_read_fast(struct bch_dev *ca, struct bch_dev_usage *usage) void bch2_dev_usage_full_read_fast(struct bch_dev *ca, struct bch_dev_usage_full *usage) { memset(usage, 0, sizeof(*usage)); - acc_u64s_percpu((u64 *) usage, (u64 __percpu *) ca->usage, dev_usage_u64s()); + acc_u64s_percpu((u64 *) usage, (u64 __percpu *) ca->usage, + sizeof(struct bch_dev_usage_full) / sizeof(u64)); } static u64 reserve_factor(u64 r) @@ -603,6 +604,13 @@ static int bch2_trigger_pointer(struct btree_trans *trans, } struct bpos bucket = PTR_BUCKET_POS(ca, &p.ptr); + if (!bucket_valid(ca, bucket.offset)) { + if (insert) { + bch2_dev_bucket_missing(ca, bucket.offset); + ret = -BCH_ERR_trigger_pointer; + } + goto err; + } if (flags & BTREE_TRIGGER_transactional) { struct bkey_i_alloc_v4 *a = bch2_trans_start_alloc_update(trans, bucket, 0); @@ -1306,13 +1314,11 @@ int bch2_dev_buckets_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets) old_bucket_gens = rcu_dereference_protected(ca->bucket_gens, 1); if (resize) { - bucket_gens->nbuckets = min(bucket_gens->nbuckets, - old_bucket_gens->nbuckets); - bucket_gens->nbuckets_minus_first = - bucket_gens->nbuckets - bucket_gens->first_bucket; + u64 copy = min(bucket_gens->nbuckets, + old_bucket_gens->nbuckets); memcpy(bucket_gens->b, old_bucket_gens->b, - bucket_gens->nbuckets); + sizeof(bucket_gens->b[0]) * copy); } rcu_assign_pointer(ca->bucket_gens, bucket_gens); diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h index 1c38b165f48b..af1532de4a37 100644 --- a/fs/bcachefs/buckets.h +++ b/fs/bcachefs/buckets.h @@ -44,6 +44,7 @@ static inline void bucket_unlock(struct bucket *b) BUILD_BUG_ON(!((union ulong_byte_assert) { .ulong = 1UL << BUCKET_LOCK_BITNR }).byte); clear_bit_unlock(BUCKET_LOCK_BITNR, (void *) &b->lock); + smp_mb__after_atomic(); wake_up_bit((void *) &b->lock, BUCKET_LOCK_BITNR); } @@ -242,11 +243,6 @@ static inline u64 dev_buckets_available(struct bch_dev *ca, /* Filesystem usage: */ -static inline unsigned dev_usage_u64s(void) -{ - return sizeof(struct bch_dev_usage) / sizeof(u64); -} - struct bch_fs_usage_short bch2_fs_usage_read_short(struct bch_fs *); diff --git a/fs/bcachefs/checksum.c b/fs/bcachefs/checksum.c index 3726689093e3..d0a34a097b80 100644 --- a/fs/bcachefs/checksum.c +++ b/fs/bcachefs/checksum.c @@ -7,17 +7,12 @@ #include "super-io.h" #include <linux/crc32c.h> -#include <linux/crypto.h> #include <linux/xxhash.h> #include <linux/key.h> #include <linux/random.h> #include <linux/ratelimit.h> -#include <linux/scatterlist.h> -#include <crypto/algapi.h> #include <crypto/chacha.h> -#include <crypto/hash.h> #include <crypto/poly1305.h> -#include <crypto/skcipher.h> #include <keys/user-type.h> /* @@ -96,116 +91,40 @@ static void bch2_checksum_update(struct bch2_checksum_state *state, const void * } } -static inline int do_encrypt_sg(struct crypto_sync_skcipher *tfm, - struct nonce nonce, - struct scatterlist *sg, size_t len) +static void bch2_chacha20_init(u32 state[CHACHA_STATE_WORDS], + const struct bch_key *key, struct nonce nonce) { - SYNC_SKCIPHER_REQUEST_ON_STACK(req, tfm); + u32 key_words[CHACHA_KEY_SIZE / sizeof(u32)]; - skcipher_request_set_sync_tfm(req, tfm); - skcipher_request_set_callback(req, 0, NULL, NULL); - skcipher_request_set_crypt(req, sg, sg, len, nonce.d); + BUILD_BUG_ON(sizeof(key_words) != sizeof(*key)); + memcpy(key_words, key, sizeof(key_words)); + le32_to_cpu_array(key_words, ARRAY_SIZE(key_words)); - int ret = crypto_skcipher_encrypt(req); - if (ret) - pr_err("got error %i from crypto_skcipher_encrypt()", ret); - - return ret; -} - -static inline int do_encrypt(struct crypto_sync_skcipher *tfm, - struct nonce nonce, - void *buf, size_t len) -{ - if (!is_vmalloc_addr(buf)) { - struct scatterlist sg = {}; - - sg_mark_end(&sg); - sg_set_page(&sg, virt_to_page(buf), len, offset_in_page(buf)); - return do_encrypt_sg(tfm, nonce, &sg, len); - } else { - DARRAY_PREALLOCATED(struct scatterlist, 4) sgl; - size_t sgl_len = 0; - int ret; - - darray_init(&sgl); - - while (len) { - unsigned offset = offset_in_page(buf); - struct scatterlist sg = { - .page_link = (unsigned long) vmalloc_to_page(buf), - .offset = offset, - .length = min(len, PAGE_SIZE - offset), - }; + BUILD_BUG_ON(sizeof(nonce) != CHACHA_IV_SIZE); + chacha_init(state, key_words, (const u8 *)nonce.d); - if (darray_push(&sgl, sg)) { - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(tfm, nonce, sgl.data, sgl_len); - if (ret) - goto err; - - nonce = nonce_add(nonce, sgl_len); - sgl_len = 0; - sgl.nr = 0; - BUG_ON(darray_push(&sgl, sg)); - } - - buf += sg.length; - len -= sg.length; - sgl_len += sg.length; - } - - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(tfm, nonce, sgl.data, sgl_len); -err: - darray_exit(&sgl); - return ret; - } + memzero_explicit(key_words, sizeof(key_words)); } -int bch2_chacha_encrypt_key(struct bch_key *key, struct nonce nonce, - void *buf, size_t len) +static void bch2_chacha20(const struct bch_key *key, struct nonce nonce, + void *data, size_t len) { - struct crypto_sync_skcipher *chacha20 = - crypto_alloc_sync_skcipher("chacha20", 0, 0); - int ret; - - ret = PTR_ERR_OR_ZERO(chacha20); - if (ret) { - pr_err("error requesting chacha20 cipher: %s", bch2_err_str(ret)); - return ret; - } - - ret = crypto_skcipher_setkey(&chacha20->base, - (void *) key, sizeof(*key)); - if (ret) { - pr_err("error from crypto_skcipher_setkey(): %s", bch2_err_str(ret)); - goto err; - } + u32 state[CHACHA_STATE_WORDS]; - ret = do_encrypt(chacha20, nonce, buf, len); -err: - crypto_free_sync_skcipher(chacha20); - return ret; + bch2_chacha20_init(state, key, nonce); + chacha20_crypt(state, data, data, len); + memzero_explicit(state, sizeof(state)); } -static int gen_poly_key(struct bch_fs *c, struct shash_desc *desc, - struct nonce nonce) +static void bch2_poly1305_init(struct poly1305_desc_ctx *desc, + struct bch_fs *c, struct nonce nonce) { - u8 key[POLY1305_KEY_SIZE]; - int ret; + u8 key[POLY1305_KEY_SIZE] = { 0 }; nonce.d[3] ^= BCH_NONCE_POLY; - memset(key, 0, sizeof(key)); - ret = do_encrypt(c->chacha20, nonce, key, sizeof(key)); - if (ret) - return ret; - - desc->tfm = c->poly1305; - crypto_shash_init(desc); - crypto_shash_update(desc, key, sizeof(key)); - return 0; + bch2_chacha20(&c->chacha20_key, nonce, key, sizeof(key)); + poly1305_init(desc, key); } struct bch_csum bch2_checksum(struct bch_fs *c, unsigned type, @@ -230,14 +149,13 @@ struct bch_csum bch2_checksum(struct bch_fs *c, unsigned type, case BCH_CSUM_chacha20_poly1305_80: case BCH_CSUM_chacha20_poly1305_128: { - SHASH_DESC_ON_STACK(desc, c->poly1305); + struct poly1305_desc_ctx dctx; u8 digest[POLY1305_DIGEST_SIZE]; struct bch_csum ret = { 0 }; - gen_poly_key(c, desc, nonce); - - crypto_shash_update(desc, data, len); - crypto_shash_final(desc, digest); + bch2_poly1305_init(&dctx, c, nonce); + poly1305_update(&dctx, data, len); + poly1305_final(&dctx, digest); memcpy(&ret, digest, bch_crc_bytes[type]); return ret; @@ -253,11 +171,12 @@ int bch2_encrypt(struct bch_fs *c, unsigned type, if (!bch2_csum_type_is_encryption(type)) return 0; - if (bch2_fs_inconsistent_on(!c->chacha20, + if (bch2_fs_inconsistent_on(!c->chacha20_key_set, c, "attempting to encrypt without encryption key")) return -BCH_ERR_no_encryption_key; - return do_encrypt(c->chacha20, nonce, data, len); + bch2_chacha20(&c->chacha20_key, nonce, data, len); + return 0; } static struct bch_csum __bch2_checksum_bio(struct bch_fs *c, unsigned type, @@ -296,26 +215,26 @@ static struct bch_csum __bch2_checksum_bio(struct bch_fs *c, unsigned type, case BCH_CSUM_chacha20_poly1305_80: case BCH_CSUM_chacha20_poly1305_128: { - SHASH_DESC_ON_STACK(desc, c->poly1305); + struct poly1305_desc_ctx dctx; u8 digest[POLY1305_DIGEST_SIZE]; struct bch_csum ret = { 0 }; - gen_poly_key(c, desc, nonce); + bch2_poly1305_init(&dctx, c, nonce); #ifdef CONFIG_HIGHMEM __bio_for_each_segment(bv, bio, *iter, *iter) { void *p = kmap_local_page(bv.bv_page) + bv.bv_offset; - crypto_shash_update(desc, p, bv.bv_len); + poly1305_update(&dctx, p, bv.bv_len); kunmap_local(p); } #else __bio_for_each_bvec(bv, bio, *iter, *iter) - crypto_shash_update(desc, + poly1305_update(&dctx, page_address(bv.bv_page) + bv.bv_offset, bv.bv_len); #endif - crypto_shash_final(desc, digest); + poly1305_final(&dctx, digest); memcpy(&ret, digest, bch_crc_bytes[type]); return ret; @@ -338,43 +257,33 @@ int __bch2_encrypt_bio(struct bch_fs *c, unsigned type, { struct bio_vec bv; struct bvec_iter iter; - DARRAY_PREALLOCATED(struct scatterlist, 4) sgl; - size_t sgl_len = 0; + u32 chacha_state[CHACHA_STATE_WORDS]; int ret = 0; - if (bch2_fs_inconsistent_on(!c->chacha20, + if (bch2_fs_inconsistent_on(!c->chacha20_key_set, c, "attempting to encrypt without encryption key")) return -BCH_ERR_no_encryption_key; - darray_init(&sgl); + bch2_chacha20_init(chacha_state, &c->chacha20_key, nonce); bio_for_each_segment(bv, bio, iter) { - struct scatterlist sg = { - .page_link = (unsigned long) bv.bv_page, - .offset = bv.bv_offset, - .length = bv.bv_len, - }; - - if (darray_push(&sgl, sg)) { - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(c->chacha20, nonce, sgl.data, sgl_len); - if (ret) - goto err; - - nonce = nonce_add(nonce, sgl_len); - sgl_len = 0; - sgl.nr = 0; - - BUG_ON(darray_push(&sgl, sg)); + void *p; + + /* + * chacha_crypt() assumes that the length is a multiple of + * CHACHA_BLOCK_SIZE on any non-final call. + */ + if (!IS_ALIGNED(bv.bv_len, CHACHA_BLOCK_SIZE)) { + bch_err_ratelimited(c, "bio not aligned for encryption"); + ret = -EIO; + break; } - sgl_len += sg.length; + p = bvec_kmap_local(&bv); + chacha20_crypt(chacha_state, p, p, bv.bv_len); + kunmap_local(p); } - - sg_mark_end(&darray_last(sgl)); - ret = do_encrypt_sg(c->chacha20, nonce, sgl.data, sgl_len); -err: - darray_exit(&sgl); + memzero_explicit(chacha_state, sizeof(chacha_state)); return ret; } @@ -650,10 +559,7 @@ int bch2_decrypt_sb_key(struct bch_fs *c, } /* decrypt real key: */ - ret = bch2_chacha_encrypt_key(&user_key, bch2_sb_key_nonce(c), - &sb_key, sizeof(sb_key)); - if (ret) - goto err; + bch2_chacha20(&user_key, bch2_sb_key_nonce(c), &sb_key, sizeof(sb_key)); if (bch2_key_is_encrypted(&sb_key)) { bch_err(c, "incorrect encryption key"); @@ -668,31 +574,6 @@ err: return ret; } -static int bch2_alloc_ciphers(struct bch_fs *c) -{ - if (c->chacha20) - return 0; - - struct crypto_sync_skcipher *chacha20 = crypto_alloc_sync_skcipher("chacha20", 0, 0); - int ret = PTR_ERR_OR_ZERO(chacha20); - if (ret) { - bch_err(c, "error requesting chacha20 module: %s", bch2_err_str(ret)); - return ret; - } - - struct crypto_shash *poly1305 = crypto_alloc_shash("poly1305", 0, 0); - ret = PTR_ERR_OR_ZERO(poly1305); - if (ret) { - bch_err(c, "error requesting poly1305 module: %s", bch2_err_str(ret)); - crypto_free_sync_skcipher(chacha20); - return ret; - } - - c->chacha20 = chacha20; - c->poly1305 = poly1305; - return 0; -} - #if 0 /* @@ -797,35 +678,21 @@ err: void bch2_fs_encryption_exit(struct bch_fs *c) { - if (c->poly1305) - crypto_free_shash(c->poly1305); - if (c->chacha20) - crypto_free_sync_skcipher(c->chacha20); + memzero_explicit(&c->chacha20_key, sizeof(c->chacha20_key)); } int bch2_fs_encryption_init(struct bch_fs *c) { struct bch_sb_field_crypt *crypt; - struct bch_key key; - int ret = 0; + int ret; crypt = bch2_sb_field_get(c->disk_sb.sb, crypt); if (!crypt) - goto out; + return 0; - ret = bch2_alloc_ciphers(c); + ret = bch2_decrypt_sb_key(c, crypt, &c->chacha20_key); if (ret) - goto out; - - ret = bch2_decrypt_sb_key(c, crypt, &key); - if (ret) - goto out; - - ret = crypto_skcipher_setkey(&c->chacha20->base, - (void *) &key.key, sizeof(key.key)); - if (ret) - goto out; -out: - memzero_explicit(&key, sizeof(key)); - return ret; + return ret; + c->chacha20_key_set = true; + return 0; } diff --git a/fs/bcachefs/checksum.h b/fs/bcachefs/checksum.h index 4ac251c8fcd8..1310782d3ae9 100644 --- a/fs/bcachefs/checksum.h +++ b/fs/bcachefs/checksum.h @@ -69,7 +69,6 @@ static inline void bch2_csum_err_msg(struct printbuf *out, bch2_csum_to_text(out, type, expected); } -int bch2_chacha_encrypt_key(struct bch_key *, struct nonce, void *, size_t); int bch2_request_key(struct bch_sb *, struct bch_key *); #ifndef __KERNEL__ int bch2_revoke_key(struct bch_sb *); @@ -156,7 +155,7 @@ static inline bool bch2_checksum_type_valid(const struct bch_fs *c, if (type >= BCH_CSUM_NR) return false; - if (bch2_csum_type_is_encryption(type) && !c->chacha20) + if (bch2_csum_type_is_encryption(type) && !c->chacha20_key_set) return false; return true; diff --git a/fs/bcachefs/clock.c b/fs/bcachefs/clock.c index 1f8e035d7119..d6dd12d74d4f 100644 --- a/fs/bcachefs/clock.c +++ b/fs/bcachefs/clock.c @@ -121,7 +121,7 @@ void bch2_kthread_io_clock_wait(struct io_clock *clock, } while (0); __set_current_state(TASK_RUNNING); - del_timer_sync(&wait.cpu_timer); + timer_delete_sync(&wait.cpu_timer); destroy_timer_on_stack(&wait.cpu_timer); bch2_io_timer_del(clock, &wait.io_timer); } diff --git a/fs/bcachefs/data_update.c b/fs/bcachefs/data_update.c index de02ebf847ec..b211c97238ab 100644 --- a/fs/bcachefs/data_update.c +++ b/fs/bcachefs/data_update.c @@ -607,7 +607,7 @@ void bch2_data_update_inflight_to_text(struct printbuf *out, struct data_update prt_newline(out); printbuf_indent_add(out, 2); bch2_data_update_opts_to_text(out, m->op.c, &m->op.opts, &m->data_opts); - prt_printf(out, "read_done:\t\%u\n", m->read_done); + prt_printf(out, "read_done:\t%u\n", m->read_done); bch2_write_op_to_text(out, &m->op); printbuf_indent_sub(out, 2); } diff --git a/fs/bcachefs/dirent.c b/fs/bcachefs/dirent.c index bf53a029f356..8a680e52c1ed 100644 --- a/fs/bcachefs/dirent.c +++ b/fs/bcachefs/dirent.c @@ -13,8 +13,8 @@ #include <linux/dcache.h> -static int bch2_casefold(struct btree_trans *trans, const struct bch_hash_info *info, - const struct qstr *str, struct qstr *out_cf) +int bch2_casefold(struct btree_trans *trans, const struct bch_hash_info *info, + const struct qstr *str, struct qstr *out_cf) { *out_cf = (struct qstr) QSTR_INIT(NULL, 0); @@ -35,18 +35,6 @@ static int bch2_casefold(struct btree_trans *trans, const struct bch_hash_info * #endif } -static inline int bch2_maybe_casefold(struct btree_trans *trans, - const struct bch_hash_info *info, - const struct qstr *str, struct qstr *out_cf) -{ - if (likely(!info->cf_encoding)) { - *out_cf = *str; - return 0; - } else { - return bch2_casefold(trans, info, str, out_cf); - } -} - static unsigned bch2_dirent_name_bytes(struct bkey_s_c_dirent d) { if (bkey_val_bytes(d.k) < offsetof(struct bch_dirent, d_name)) @@ -287,8 +275,8 @@ static void dirent_init_casefolded_name(struct bkey_i_dirent *dirent, EBUG_ON(!dirent->v.d_casefold); EBUG_ON(!cf_name->len); - dirent->v.d_cf_name_block.d_name_len = name->len; - dirent->v.d_cf_name_block.d_cf_name_len = cf_name->len; + dirent->v.d_cf_name_block.d_name_len = cpu_to_le16(name->len); + dirent->v.d_cf_name_block.d_cf_name_len = cpu_to_le16(cf_name->len); memcpy(&dirent->v.d_cf_name_block.d_names[0], name->name, name->len); memcpy(&dirent->v.d_cf_name_block.d_names[name->len], cf_name->name, cf_name->len); memset(&dirent->v.d_cf_name_block.d_names[name->len + cf_name->len], 0, @@ -697,7 +685,7 @@ static int bch2_dir_emit(struct dir_context *ctx, struct bkey_s_c_dirent d, subv vfs_d_type(d.v->d_type)); if (ret) ctx->pos = d.k->p.offset + 1; - return ret; + return !ret; } int bch2_readdir(struct bch_fs *c, subvol_inum inum, struct dir_context *ctx) @@ -722,7 +710,7 @@ int bch2_readdir(struct bch_fs *c, subvol_inum inum, struct dir_context *ctx) if (ret2 > 0) continue; - ret2 ?: drop_locks_do(trans, bch2_dir_emit(ctx, dirent, target)); + ret2 ?: (bch2_trans_unlock(trans), bch2_dir_emit(ctx, dirent, target)); }))); bch2_bkey_buf_exit(&sk, c); diff --git a/fs/bcachefs/dirent.h b/fs/bcachefs/dirent.h index 0880772b80a9..9838a7ba7ed1 100644 --- a/fs/bcachefs/dirent.h +++ b/fs/bcachefs/dirent.h @@ -23,6 +23,21 @@ struct bch_fs; struct bch_hash_info; struct bch_inode_info; +int bch2_casefold(struct btree_trans *, const struct bch_hash_info *, + const struct qstr *, struct qstr *); + +static inline int bch2_maybe_casefold(struct btree_trans *trans, + const struct bch_hash_info *info, + const struct qstr *str, struct qstr *out_cf) +{ + if (likely(!info->cf_encoding)) { + *out_cf = *str; + return 0; + } else { + return bch2_casefold(trans, info, str, out_cf); + } +} + struct qstr bch2_dirent_get_name(struct bkey_s_c_dirent d); static inline unsigned dirent_val_u64s(unsigned len, unsigned cf_len) diff --git a/fs/bcachefs/disk_groups.c b/fs/bcachefs/disk_groups.c index 1186280b29e9..2ca3cbf12b71 100644 --- a/fs/bcachefs/disk_groups.c +++ b/fs/bcachefs/disk_groups.c @@ -470,23 +470,22 @@ inval: int __bch2_dev_group_set(struct bch_fs *c, struct bch_dev *ca, const char *name) { - struct bch_member *mi; - int ret, v = -1; + lockdep_assert_held(&c->sb_lock); - if (!strlen(name) || !strcmp(name, "none")) - return 0; - v = bch2_disk_path_find_or_create(&c->disk_sb, name); - if (v < 0) - return v; + if (!strlen(name) || !strcmp(name, "none")) { + struct bch_member *mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx); + SET_BCH_MEMBER_GROUP(mi, 0); + } else { + int v = bch2_disk_path_find_or_create(&c->disk_sb, name); + if (v < 0) + return v; - ret = bch2_sb_disk_groups_to_cpu(c); - if (ret) - return ret; + struct bch_member *mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx); + SET_BCH_MEMBER_GROUP(mi, v + 1); + } - mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx); - SET_BCH_MEMBER_GROUP(mi, v + 1); - return 0; + return bch2_sb_disk_groups_to_cpu(c); } int bch2_dev_group_set(struct bch_fs *c, struct bch_dev *ca, const char *name) diff --git a/fs/bcachefs/ec.c b/fs/bcachefs/ec.c index a396865e8b17..fff58b78327c 100644 --- a/fs/bcachefs/ec.c +++ b/fs/bcachefs/ec.c @@ -2204,10 +2204,10 @@ void bch2_fs_ec_stop(struct bch_fs *c) static bool bch2_fs_ec_flush_done(struct bch_fs *c) { - bool ret; + sched_annotate_sleep(); mutex_lock(&c->ec_stripe_new_lock); - ret = list_empty(&c->ec_stripe_new_list); + bool ret = list_empty(&c->ec_stripe_new_list); mutex_unlock(&c->ec_stripe_new_lock); return ret; diff --git a/fs/bcachefs/ec.h b/fs/bcachefs/ec.h index 62d27e04d763..51893e1ee874 100644 --- a/fs/bcachefs/ec.h +++ b/fs/bcachefs/ec.h @@ -160,6 +160,7 @@ static inline void gc_stripe_unlock(struct gc_stripe *s) BUILD_BUG_ON(!((union ulong_byte_assert) { .ulong = 1UL << BUCKET_LOCK_BITNR }).byte); clear_bit_unlock(BUCKET_LOCK_BITNR, (void *) &s->lock); + smp_mb__after_atomic(); wake_up_bit((void *) &s->lock, BUCKET_LOCK_BITNR); } diff --git a/fs/bcachefs/errcode.h b/fs/bcachefs/errcode.h index c8696f01eb14..d9ebffa5b3a2 100644 --- a/fs/bcachefs/errcode.h +++ b/fs/bcachefs/errcode.h @@ -269,7 +269,7 @@ x(BCH_ERR_invalid_sb, invalid_sb_downgrade) \ x(BCH_ERR_invalid, invalid_bkey) \ x(BCH_ERR_operation_blocked, nocow_lock_blocked) \ - x(EIO, journal_shutdown) \ + x(EROFS, journal_shutdown) \ x(EIO, journal_flush_err) \ x(EIO, journal_write_err) \ x(EIO, btree_node_read_err) \ @@ -287,7 +287,7 @@ x(EIO, mark_stripe) \ x(EIO, stripe_reconstruct) \ x(EIO, key_type_error) \ - x(EIO, extent_poisened) \ + x(EIO, extent_poisoned) \ x(EIO, missing_indirect_extent) \ x(EIO, invalidate_stripe_to_dev) \ x(EIO, no_encryption_key) \ diff --git a/fs/bcachefs/error.c b/fs/bcachefs/error.c index baf5dfb32298..6b8695b1349c 100644 --- a/fs/bcachefs/error.c +++ b/fs/bcachefs/error.c @@ -272,9 +272,6 @@ static struct fsck_err_state *fsck_err_get(struct bch_fs *c, { struct fsck_err_state *s; - if (!test_bit(BCH_FS_fsck_running, &c->flags)) - return NULL; - list_for_each_entry(s, &c->fsck_error_msgs, list) if (s->id == id) { /* @@ -481,7 +478,9 @@ int __bch2_fsck_err(struct bch_fs *c, } else if (!test_bit(BCH_FS_fsck_running, &c->flags)) { if (c->opts.errors != BCH_ON_ERROR_continue || !(flags & (FSCK_CAN_FIX|FSCK_CAN_IGNORE))) { - prt_str(out, ", shutting down"); + prt_str_indented(out, ", shutting down\n" + "error not marked as autofix and not in fsck\n" + "run fsck, and forward to devs so error can be marked for self-healing"); inconsistent = true; print = true; ret = -BCH_ERR_fsck_errors_not_fixed; @@ -639,14 +638,14 @@ int __bch2_bkey_fsck_err(struct bch_fs *c, return ret; } -void bch2_flush_fsck_errs(struct bch_fs *c) +static void __bch2_flush_fsck_errs(struct bch_fs *c, bool print) { struct fsck_err_state *s, *n; mutex_lock(&c->fsck_error_msgs_lock); list_for_each_entry_safe(s, n, &c->fsck_error_msgs, list) { - if (s->ratelimited && s->last_msg) + if (print && s->ratelimited && s->last_msg) bch_err(c, "Saw %llu errors like:\n %s", s->nr, s->last_msg); list_del(&s->list); @@ -657,6 +656,16 @@ void bch2_flush_fsck_errs(struct bch_fs *c) mutex_unlock(&c->fsck_error_msgs_lock); } +void bch2_flush_fsck_errs(struct bch_fs *c) +{ + __bch2_flush_fsck_errs(c, true); +} + +void bch2_free_fsck_errs(struct bch_fs *c) +{ + __bch2_flush_fsck_errs(c, false); +} + int bch2_inum_offset_err_msg_trans(struct btree_trans *trans, struct printbuf *out, subvol_inum inum, u64 offset) { diff --git a/fs/bcachefs/error.h b/fs/bcachefs/error.h index d0d024dc714b..4a364fd44abe 100644 --- a/fs/bcachefs/error.h +++ b/fs/bcachefs/error.h @@ -93,6 +93,7 @@ int __bch2_fsck_err(struct bch_fs *, struct btree_trans *, _flags, BCH_FSCK_ERR_##_err_type, __VA_ARGS__) void bch2_flush_fsck_errs(struct bch_fs *); +void bch2_free_fsck_errs(struct bch_fs *); #define fsck_err_wrap(_do) \ ({ \ diff --git a/fs/bcachefs/extents.c b/fs/bcachefs/extents.c index ae7c7a177e10..e597fb9c9823 100644 --- a/fs/bcachefs/extents.c +++ b/fs/bcachefs/extents.c @@ -139,7 +139,7 @@ int bch2_bkey_pick_read_device(struct bch_fs *c, struct bkey_s_c k, struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); if (bch2_bkey_extent_ptrs_flags(ptrs) & BIT_ULL(BCH_EXTENT_FLAG_poisoned)) - return -BCH_ERR_extent_poisened; + return -BCH_ERR_extent_poisoned; rcu_read_lock(); const union bch_extent_entry *entry; @@ -1056,8 +1056,9 @@ bch2_extent_has_ptr(struct bkey_s_c k1, struct extent_ptr_decoded p1, struct bke static bool want_cached_ptr(struct bch_fs *c, struct bch_io_opts *opts, struct bch_extent_ptr *ptr) { - if (!opts->promote_target || - !bch2_dev_in_target(c, ptr->dev, opts->promote_target)) + unsigned target = opts->promote_target ?: opts->foreground_target; + + if (target && !bch2_dev_in_target(c, ptr->dev, target)) return false; struct bch_dev *ca = bch2_dev_rcu_noerror(c, ptr->dev); diff --git a/fs/bcachefs/fs-io-buffered.c b/fs/bcachefs/fs-io-buffered.c index 19d4599918dc..e3a75dcca60c 100644 --- a/fs/bcachefs/fs-io-buffered.c +++ b/fs/bcachefs/fs-io-buffered.c @@ -225,11 +225,26 @@ static void bchfs_read(struct btree_trans *trans, bch2_read_extent(trans, rbio, iter.pos, data_btree, k, offset_into_extent, flags); - swap(rbio->bio.bi_iter.bi_size, bytes); + /* + * Careful there's a landmine here if bch2_read_extent() ever + * starts returning transaction restarts here. + * + * We've changed rbio->bi_iter.bi_size to be "bytes we can read + * from this extent" with the swap call, and we restore it + * below. That restore needs to come before checking for + * errors. + * + * But unlike __bch2_read(), we use the rbio bvec iter, not one + * on the stack, so we can't do the restore right after the + * bch2_read_extent() call: we don't own that iterator anymore + * if BCH_READ_last_fragment is set, since we may have submitted + * that rbio instead of cloning it. + */ if (flags & BCH_READ_last_fragment) break; + swap(rbio->bio.bi_iter.bi_size, bytes); bio_advance(&rbio->bio, bytes); err: if (ret && diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c index 65c2c33d253d..9657144666b8 100644 --- a/fs/bcachefs/fs-io.c +++ b/fs/bcachefs/fs-io.c @@ -144,10 +144,25 @@ int __must_check bch2_write_inode_size(struct bch_fs *c, void __bch2_i_sectors_acct(struct bch_fs *c, struct bch_inode_info *inode, struct quota_res *quota_res, s64 sectors) { - bch2_fs_inconsistent_on((s64) inode->v.i_blocks + sectors < 0, c, - "inode %lu i_blocks underflow: %llu + %lli < 0 (ondisk %lli)", - inode->v.i_ino, (u64) inode->v.i_blocks, sectors, - inode->ei_inode.bi_sectors); + if (unlikely((s64) inode->v.i_blocks + sectors < 0)) { + struct printbuf buf = PRINTBUF; + bch2_log_msg_start(c, &buf); + prt_printf(&buf, "inode %lu i_blocks underflow: %llu + %lli < 0 (ondisk %lli)", + inode->v.i_ino, (u64) inode->v.i_blocks, sectors, + inode->ei_inode.bi_sectors); + + bool repeat = false, print = false, suppress = false; + bch2_count_fsck_err(c, vfs_inode_i_blocks_underflow, buf.buf, &repeat, &print, &suppress); + if (print) + bch2_print_str(c, buf.buf); + printbuf_exit(&buf); + + if (sectors < 0) + sectors = -inode->v.i_blocks; + else + sectors = 0; + } + inode->v.i_blocks += sectors; #ifdef CONFIG_BCACHEFS_QUOTA @@ -502,11 +517,22 @@ int bchfs_truncate(struct mnt_idmap *idmap, goto err; } - bch2_fs_inconsistent_on(!inode->v.i_size && inode->v.i_blocks && - !bch2_journal_error(&c->journal), c, - "inode %lu truncated to 0 but i_blocks %llu (ondisk %lli)", - inode->v.i_ino, (u64) inode->v.i_blocks, - inode->ei_inode.bi_sectors); + if (unlikely(!inode->v.i_size && inode->v.i_blocks && + !bch2_journal_error(&c->journal))) { + struct printbuf buf = PRINTBUF; + bch2_log_msg_start(c, &buf); + prt_printf(&buf, + "inode %lu truncated to 0 but i_blocks %llu (ondisk %lli)", + inode->v.i_ino, (u64) inode->v.i_blocks, + inode->ei_inode.bi_sectors); + + bool repeat = false, print = false, suppress = false; + bch2_count_fsck_err(c, vfs_inode_i_blocks_not_zero_at_truncate, buf.buf, + &repeat, &print, &suppress); + if (print) + bch2_print_str(c, buf.buf); + printbuf_exit(&buf); + } ret = bch2_setattr_nonsize(idmap, inode, iattr); err: diff --git a/fs/bcachefs/fs-ioctl.c b/fs/bcachefs/fs-ioctl.c index c1553e44e049..a82dfce9e4ad 100644 --- a/fs/bcachefs/fs-ioctl.c +++ b/fs/bcachefs/fs-ioctl.c @@ -21,206 +21,6 @@ #define FSOP_GOING_FLAGS_LOGFLUSH 0x1 /* flush log but not data */ #define FSOP_GOING_FLAGS_NOLOGFLUSH 0x2 /* don't flush log nor data */ -struct flags_set { - unsigned mask; - unsigned flags; - - unsigned projid; - - bool set_projinherit; - bool projinherit; -}; - -static int bch2_inode_flags_set(struct btree_trans *trans, - struct bch_inode_info *inode, - struct bch_inode_unpacked *bi, - void *p) -{ - struct bch_fs *c = inode->v.i_sb->s_fs_info; - /* - * We're relying on btree locking here for exclusion with other ioctl - * calls - use the flags in the btree (@bi), not inode->i_flags: - */ - struct flags_set *s = p; - unsigned newflags = s->flags; - unsigned oldflags = bi->bi_flags & s->mask; - - if (((newflags ^ oldflags) & (BCH_INODE_append|BCH_INODE_immutable)) && - !capable(CAP_LINUX_IMMUTABLE)) - return -EPERM; - - if (!S_ISREG(bi->bi_mode) && - !S_ISDIR(bi->bi_mode) && - (newflags & (BCH_INODE_nodump|BCH_INODE_noatime)) != newflags) - return -EINVAL; - - if ((newflags ^ oldflags) & BCH_INODE_casefolded) { -#ifdef CONFIG_UNICODE - int ret = 0; - /* Not supported on individual files. */ - if (!S_ISDIR(bi->bi_mode)) - return -EOPNOTSUPP; - - /* - * Make sure the dir is empty, as otherwise we'd need to - * rehash everything and update the dirent keys. - */ - ret = bch2_empty_dir_trans(trans, inode_inum(inode)); - if (ret < 0) - return ret; - - ret = bch2_request_incompat_feature(c,bcachefs_metadata_version_casefolding); - if (ret) - return ret; - - bch2_check_set_feature(c, BCH_FEATURE_casefolding); -#else - printk(KERN_ERR "Cannot use casefolding on a kernel without CONFIG_UNICODE\n"); - return -EOPNOTSUPP; -#endif - } - - if (s->set_projinherit) { - bi->bi_fields_set &= ~(1 << Inode_opt_project); - bi->bi_fields_set |= ((int) s->projinherit << Inode_opt_project); - } - - bi->bi_flags &= ~s->mask; - bi->bi_flags |= newflags; - - bi->bi_ctime = timespec_to_bch2_time(c, current_time(&inode->v)); - return 0; -} - -static int bch2_ioc_getflags(struct bch_inode_info *inode, int __user *arg) -{ - unsigned flags = map_flags(bch_flags_to_uflags, inode->ei_inode.bi_flags); - - return put_user(flags, arg); -} - -static int bch2_ioc_setflags(struct bch_fs *c, - struct file *file, - struct bch_inode_info *inode, - void __user *arg) -{ - struct flags_set s = { .mask = map_defined(bch_flags_to_uflags) }; - unsigned uflags; - int ret; - - if (get_user(uflags, (int __user *) arg)) - return -EFAULT; - - s.flags = map_flags_rev(bch_flags_to_uflags, uflags); - if (uflags) - return -EOPNOTSUPP; - - ret = mnt_want_write_file(file); - if (ret) - return ret; - - inode_lock(&inode->v); - if (!inode_owner_or_capable(file_mnt_idmap(file), &inode->v)) { - ret = -EACCES; - goto setflags_out; - } - - mutex_lock(&inode->ei_update_lock); - ret = bch2_subvol_is_ro(c, inode->ei_inum.subvol) ?: - bch2_write_inode(c, inode, bch2_inode_flags_set, &s, - ATTR_CTIME); - mutex_unlock(&inode->ei_update_lock); - -setflags_out: - inode_unlock(&inode->v); - mnt_drop_write_file(file); - return ret; -} - -static int bch2_ioc_fsgetxattr(struct bch_inode_info *inode, - struct fsxattr __user *arg) -{ - struct fsxattr fa = { 0 }; - - fa.fsx_xflags = map_flags(bch_flags_to_xflags, inode->ei_inode.bi_flags); - - if (inode->ei_inode.bi_fields_set & (1 << Inode_opt_project)) - fa.fsx_xflags |= FS_XFLAG_PROJINHERIT; - - fa.fsx_projid = inode->ei_qid.q[QTYP_PRJ]; - - if (copy_to_user(arg, &fa, sizeof(fa))) - return -EFAULT; - - return 0; -} - -static int fssetxattr_inode_update_fn(struct btree_trans *trans, - struct bch_inode_info *inode, - struct bch_inode_unpacked *bi, - void *p) -{ - struct flags_set *s = p; - - if (s->projid != bi->bi_project) { - bi->bi_fields_set |= 1U << Inode_opt_project; - bi->bi_project = s->projid; - } - - return bch2_inode_flags_set(trans, inode, bi, p); -} - -static int bch2_ioc_fssetxattr(struct bch_fs *c, - struct file *file, - struct bch_inode_info *inode, - struct fsxattr __user *arg) -{ - struct flags_set s = { .mask = map_defined(bch_flags_to_xflags) }; - struct fsxattr fa; - int ret; - - if (copy_from_user(&fa, arg, sizeof(fa))) - return -EFAULT; - - s.set_projinherit = true; - s.projinherit = (fa.fsx_xflags & FS_XFLAG_PROJINHERIT) != 0; - fa.fsx_xflags &= ~FS_XFLAG_PROJINHERIT; - - s.flags = map_flags_rev(bch_flags_to_xflags, fa.fsx_xflags); - if (fa.fsx_xflags) - return -EOPNOTSUPP; - - if (fa.fsx_projid >= U32_MAX) - return -EINVAL; - - /* - * inode fields accessible via the xattr interface are stored with a +1 - * bias, so that 0 means unset: - */ - s.projid = fa.fsx_projid + 1; - - ret = mnt_want_write_file(file); - if (ret) - return ret; - - inode_lock(&inode->v); - if (!inode_owner_or_capable(file_mnt_idmap(file), &inode->v)) { - ret = -EACCES; - goto err; - } - - mutex_lock(&inode->ei_update_lock); - ret = bch2_subvol_is_ro(c, inode->ei_inum.subvol) ?: - bch2_set_projid(c, inode, fa.fsx_projid) ?: - bch2_write_inode(c, inode, fssetxattr_inode_update_fn, &s, - ATTR_CTIME); - mutex_unlock(&inode->ei_update_lock); -err: - inode_unlock(&inode->v); - mnt_drop_write_file(file); - return ret; -} - static int bch2_reinherit_attrs_fn(struct btree_trans *trans, struct bch_inode_info *inode, struct bch_inode_unpacked *bi, @@ -558,23 +358,6 @@ long bch2_fs_file_ioctl(struct file *file, unsigned cmd, unsigned long arg) long ret; switch (cmd) { - case FS_IOC_GETFLAGS: - ret = bch2_ioc_getflags(inode, (int __user *) arg); - break; - - case FS_IOC_SETFLAGS: - ret = bch2_ioc_setflags(c, file, inode, (int __user *) arg); - break; - - case FS_IOC_FSGETXATTR: - ret = bch2_ioc_fsgetxattr(inode, (void __user *) arg); - break; - - case FS_IOC_FSSETXATTR: - ret = bch2_ioc_fssetxattr(c, file, inode, - (void __user *) arg); - break; - case BCHFS_IOC_REINHERIT_ATTRS: ret = bch2_ioc_reinherit_attrs(c, file, inode, (void __user *) arg); diff --git a/fs/bcachefs/fs-ioctl.h b/fs/bcachefs/fs-ioctl.h index ecd3bfdcde21..a657e4994b71 100644 --- a/fs/bcachefs/fs-ioctl.h +++ b/fs/bcachefs/fs-ioctl.h @@ -2,81 +2,6 @@ #ifndef _BCACHEFS_FS_IOCTL_H #define _BCACHEFS_FS_IOCTL_H -/* Inode flags: */ - -/* bcachefs inode flags -> vfs inode flags: */ -static const __maybe_unused unsigned bch_flags_to_vfs[] = { - [__BCH_INODE_sync] = S_SYNC, - [__BCH_INODE_immutable] = S_IMMUTABLE, - [__BCH_INODE_append] = S_APPEND, - [__BCH_INODE_noatime] = S_NOATIME, - [__BCH_INODE_casefolded] = S_CASEFOLD, -}; - -/* bcachefs inode flags -> FS_IOC_GETFLAGS: */ -static const __maybe_unused unsigned bch_flags_to_uflags[] = { - [__BCH_INODE_sync] = FS_SYNC_FL, - [__BCH_INODE_immutable] = FS_IMMUTABLE_FL, - [__BCH_INODE_append] = FS_APPEND_FL, - [__BCH_INODE_nodump] = FS_NODUMP_FL, - [__BCH_INODE_noatime] = FS_NOATIME_FL, - [__BCH_INODE_casefolded] = FS_CASEFOLD_FL, -}; - -/* bcachefs inode flags -> FS_IOC_FSGETXATTR: */ -static const __maybe_unused unsigned bch_flags_to_xflags[] = { - [__BCH_INODE_sync] = FS_XFLAG_SYNC, - [__BCH_INODE_immutable] = FS_XFLAG_IMMUTABLE, - [__BCH_INODE_append] = FS_XFLAG_APPEND, - [__BCH_INODE_nodump] = FS_XFLAG_NODUMP, - [__BCH_INODE_noatime] = FS_XFLAG_NOATIME, - //[__BCH_INODE_PROJINHERIT] = FS_XFLAG_PROJINHERIT; -}; - -#define set_flags(_map, _in, _out) \ -do { \ - unsigned _i; \ - \ - for (_i = 0; _i < ARRAY_SIZE(_map); _i++) \ - if ((_in) & (1 << _i)) \ - (_out) |= _map[_i]; \ - else \ - (_out) &= ~_map[_i]; \ -} while (0) - -#define map_flags(_map, _in) \ -({ \ - unsigned _out = 0; \ - \ - set_flags(_map, _in, _out); \ - _out; \ -}) - -#define map_flags_rev(_map, _in) \ -({ \ - unsigned _i, _out = 0; \ - \ - for (_i = 0; _i < ARRAY_SIZE(_map); _i++) \ - if ((_in) & _map[_i]) { \ - (_out) |= 1 << _i; \ - (_in) &= ~_map[_i]; \ - } \ - (_out); \ -}) - -#define map_defined(_map) \ -({ \ - unsigned _in = ~0; \ - \ - map_flags_rev(_map, _in); \ -}) - -/* Set VFS inode flags from bcachefs inode: */ -static inline void bch2_inode_flags_to_vfs(struct bch_inode_info *inode) -{ - set_flags(bch_flags_to_vfs, inode->ei_inode.bi_flags, inode->v.i_flags); -} - long bch2_fs_file_ioctl(struct file *, unsigned, unsigned long); long bch2_compat_fs_ioctl(struct file *, unsigned, unsigned long); diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c index 5a41b1a8e54f..b6801861c66f 100644 --- a/fs/bcachefs/fs.c +++ b/fs/bcachefs/fs.c @@ -33,6 +33,7 @@ #include <linux/backing-dev.h> #include <linux/exportfs.h> #include <linux/fiemap.h> +#include <linux/fileattr.h> #include <linux/fs_context.h> #include <linux/module.h> #include <linux/pagemap.h> @@ -51,6 +52,24 @@ static void bch2_vfs_inode_init(struct btree_trans *, subvol_inum, struct bch_inode_unpacked *, struct bch_subvolume *); +/* Set VFS inode flags from bcachefs inode: */ +static inline void bch2_inode_flags_to_vfs(struct bch_fs *c, struct bch_inode_info *inode) +{ + static const __maybe_unused unsigned bch_flags_to_vfs[] = { + [__BCH_INODE_sync] = S_SYNC, + [__BCH_INODE_immutable] = S_IMMUTABLE, + [__BCH_INODE_append] = S_APPEND, + [__BCH_INODE_noatime] = S_NOATIME, + }; + + set_flags(bch_flags_to_vfs, inode->ei_inode.bi_flags, inode->v.i_flags); + + if (bch2_inode_casefold(c, &inode->ei_inode)) + inode->v.i_flags |= S_CASEFOLD; + else + inode->v.i_flags &= ~S_CASEFOLD; +} + void bch2_inode_update_after_write(struct btree_trans *trans, struct bch_inode_info *inode, struct bch_inode_unpacked *bi, @@ -79,7 +98,7 @@ void bch2_inode_update_after_write(struct btree_trans *trans, inode->ei_inode = *bi; - bch2_inode_flags_to_vfs(inode); + bch2_inode_flags_to_vfs(c, inode); } int __must_check bch2_write_inode(struct bch_fs *c, @@ -631,13 +650,18 @@ static struct bch_inode_info *bch2_lookup_trans(struct btree_trans *trans, const struct qstr *name) { struct bch_fs *c = trans->c; - struct btree_iter dirent_iter = {}; subvol_inum inum = {}; struct printbuf buf = PRINTBUF; + struct qstr lookup_name; + int ret = bch2_maybe_casefold(trans, dir_hash_info, name, &lookup_name); + if (ret) + return ERR_PTR(ret); + + struct btree_iter dirent_iter = {}; struct bkey_s_c k = bch2_hash_lookup(trans, &dirent_iter, bch2_dirent_hash_desc, - dir_hash_info, dir, name, 0); - int ret = bkey_err(k); + dir_hash_info, dir, &lookup_name, 0); + ret = bkey_err(k); if (ret) return ERR_PTR(ret); @@ -825,6 +849,9 @@ int __bch2_unlink(struct inode *vdir, struct dentry *dentry, */ set_nlink(&inode->v, 0); } + + if (IS_CASEFOLDED(vdir)) + d_invalidate(dentry); err: bch2_trans_put(trans); bch2_unlock_inodes(INODE_UPDATE_LOCK, dir, inode); @@ -1235,10 +1262,20 @@ static int bch2_tmpfile(struct mnt_idmap *idmap, return finish_open_simple(file, 0); } +struct bch_fiemap_extent { + struct bkey_buf kbuf; + unsigned flags; +}; + static int bch2_fill_extent(struct bch_fs *c, struct fiemap_extent_info *info, - struct bkey_s_c k, unsigned flags) + struct bch_fiemap_extent *fe) { + struct bkey_s_c k = bkey_i_to_s_c(fe->kbuf.k); + unsigned flags = fe->flags; + + BUG_ON(!k.k->size); + if (bkey_extent_is_direct_data(k.k)) { struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k); const union bch_extent_entry *entry; @@ -1291,110 +1328,223 @@ static int bch2_fill_extent(struct bch_fs *c, } } -static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info, - u64 start, u64 len) +/* + * Scan a range of an inode for data in pagecache. + * + * Intended to be retryable, so don't modify the output params until success is + * imminent. + */ +static int +bch2_fiemap_hole_pagecache(struct inode *vinode, u64 *start, u64 *end, + bool nonblock) { - struct bch_fs *c = vinode->i_sb->s_fs_info; - struct bch_inode_info *ei = to_bch_ei(vinode); - struct btree_trans *trans; - struct btree_iter iter; - struct bkey_s_c k; - struct bkey_buf cur, prev; - bool have_extent = false; - int ret = 0; + loff_t dstart, dend; - ret = fiemap_prep(&ei->v, info, start, &len, FIEMAP_FLAG_SYNC); - if (ret) + dstart = bch2_seek_pagecache_data(vinode, *start, *end, 0, nonblock); + if (dstart < 0) + return dstart; + + if (dstart == *end) { + *start = dstart; + return 0; + } + + dend = bch2_seek_pagecache_hole(vinode, dstart, *end, 0, nonblock); + if (dend < 0) + return dend; + + /* race */ + BUG_ON(dstart == dend); + + *start = dstart; + *end = dend; + return 0; +} + +/* + * Scan a range of pagecache that corresponds to a file mapping hole in the + * extent btree. If data is found, fake up an extent key so it looks like a + * delalloc extent to the rest of the fiemap processing code. + */ +static int +bch2_next_fiemap_pagecache_extent(struct btree_trans *trans, struct bch_inode_info *inode, + u64 start, u64 end, struct bch_fiemap_extent *cur) +{ + struct bch_fs *c = trans->c; + struct bkey_i_extent *delextent; + struct bch_extent_ptr ptr = {}; + loff_t dstart = start << 9, dend = end << 9; + int ret; + + /* + * We hold btree locks here so we cannot block on folio locks without + * dropping trans locks first. Run a nonblocking scan for the common + * case of no folios over holes and fall back on failure. + * + * Note that dropping locks like this is technically racy against + * writeback inserting to the extent tree, but a non-sync fiemap scan is + * fundamentally racy with writeback anyways. Therefore, just report the + * range as delalloc regardless of whether we have to cycle trans locks. + */ + ret = bch2_fiemap_hole_pagecache(&inode->v, &dstart, &dend, true); + if (ret == -EAGAIN) + ret = drop_locks_do(trans, + bch2_fiemap_hole_pagecache(&inode->v, &dstart, &dend, false)); + if (ret < 0) return ret; - struct bpos end = POS(ei->v.i_ino, (start + len) >> 9); - if (start + len < start) - return -EINVAL; + /* + * Create a fake extent key in the buffer. We have to add a dummy extent + * pointer for the fill code to add an extent entry. It's explicitly + * zeroed to reflect delayed allocation (i.e. phys offset 0). + */ + bch2_bkey_buf_realloc(&cur->kbuf, c, sizeof(*delextent) / sizeof(u64)); + delextent = bkey_extent_init(cur->kbuf.k); + delextent->k.p = POS(inode->ei_inum.inum, dend >> 9); + delextent->k.size = (dend - dstart) >> 9; + bch2_bkey_append_ptr(&delextent->k_i, ptr); - start >>= 9; + cur->flags = FIEMAP_EXTENT_DELALLOC; - bch2_bkey_buf_init(&cur); - bch2_bkey_buf_init(&prev); - trans = bch2_trans_get(c); + return 0; +} + +static int bch2_next_fiemap_extent(struct btree_trans *trans, + struct bch_inode_info *inode, + u64 start, u64 end, + struct bch_fiemap_extent *cur) +{ + u32 snapshot; + int ret = bch2_subvolume_get_snapshot(trans, inode->ei_inum.subvol, &snapshot); + if (ret) + return ret; + struct btree_iter iter; bch2_trans_iter_init(trans, &iter, BTREE_ID_extents, - POS(ei->v.i_ino, start), 0); + SPOS(inode->ei_inum.inum, start, snapshot), 0); - while (!ret || bch2_err_matches(ret, BCH_ERR_transaction_restart)) { - enum btree_id data_btree = BTREE_ID_extents; + struct bkey_s_c k = + bch2_btree_iter_peek_max(trans, &iter, POS(inode->ei_inum.inum, end)); + ret = bkey_err(k); + if (ret) + goto err; - bch2_trans_begin(trans); + ret = bch2_next_fiemap_pagecache_extent(trans, inode, start, end, cur); + if (ret) + goto err; - u32 snapshot; - ret = bch2_subvolume_get_snapshot(trans, ei->ei_inum.subvol, &snapshot); - if (ret) - continue; + struct bpos pagecache_start = bkey_start_pos(&cur->kbuf.k->k); - bch2_btree_iter_set_snapshot(trans, &iter, snapshot); + /* + * Does the pagecache or the btree take precedence? + * + * It _should_ be the pagecache, so that we correctly report delalloc + * extents when dirty in the pagecache (we're COW, after all). + * + * But we'd have to add per-sector writeback tracking to + * bch_folio_state, otherwise we report delalloc extents for clean + * cached data in the pagecache. + * + * We should do this, but even then fiemap won't report stable mappings: + * on bcachefs data moves around in the background (copygc, rebalance) + * and we don't provide a way for userspace to lock that out. + */ + if (k.k && + bkey_le(bpos_max(iter.pos, bkey_start_pos(k.k)), + pagecache_start)) { + bch2_bkey_buf_reassemble(&cur->kbuf, trans->c, k); + bch2_cut_front(iter.pos, cur->kbuf.k); + bch2_cut_back(POS(inode->ei_inum.inum, end), cur->kbuf.k); + cur->flags = 0; + } else if (k.k) { + bch2_cut_back(bkey_start_pos(k.k), cur->kbuf.k); + } - k = bch2_btree_iter_peek_max(trans, &iter, end); - ret = bkey_err(k); + if (cur->kbuf.k->k.type == KEY_TYPE_reflink_p) { + unsigned sectors = cur->kbuf.k->k.size; + s64 offset_into_extent = 0; + enum btree_id data_btree = BTREE_ID_extents; + ret = bch2_read_indirect_extent(trans, &data_btree, &offset_into_extent, + &cur->kbuf); if (ret) - continue; + goto err; - if (!k.k) - break; + struct bkey_i *k = cur->kbuf.k; + sectors = min_t(unsigned, sectors, k->k.size - offset_into_extent); - if (!bkey_extent_is_data(k.k) && - k.k->type != KEY_TYPE_reservation) { - bch2_btree_iter_advance(trans, &iter); - continue; - } + bch2_cut_front(POS(k->k.p.inode, + bkey_start_offset(&k->k) + offset_into_extent), + k); + bch2_key_resize(&k->k, sectors); + k->k.p = iter.pos; + k->k.p.offset += k->k.size; + } +err: + bch2_trans_iter_exit(trans, &iter); + return ret; +} - s64 offset_into_extent = iter.pos.offset - bkey_start_offset(k.k); - unsigned sectors = k.k->size - offset_into_extent; +static int bch2_fiemap(struct inode *vinode, struct fiemap_extent_info *info, + u64 start, u64 len) +{ + struct bch_fs *c = vinode->i_sb->s_fs_info; + struct bch_inode_info *ei = to_bch_ei(vinode); + struct btree_trans *trans; + struct bch_fiemap_extent cur, prev; + int ret = 0; - bch2_bkey_buf_reassemble(&cur, c, k); + ret = fiemap_prep(&ei->v, info, start, &len, 0); + if (ret) + return ret; - ret = bch2_read_indirect_extent(trans, &data_btree, - &offset_into_extent, &cur); + if (start + len < start) + return -EINVAL; + + start >>= 9; + u64 end = (start + len) >> 9; + + bch2_bkey_buf_init(&cur.kbuf); + bch2_bkey_buf_init(&prev.kbuf); + bkey_init(&prev.kbuf.k->k); + + trans = bch2_trans_get(c); + + while (start < end) { + ret = lockrestart_do(trans, + bch2_next_fiemap_extent(trans, ei, start, end, &cur)); if (ret) - continue; + goto err; - k = bkey_i_to_s_c(cur.k); - bch2_bkey_buf_realloc(&prev, c, k.k->u64s); + BUG_ON(bkey_start_offset(&cur.kbuf.k->k) < start); + BUG_ON(cur.kbuf.k->k.p.offset > end); - sectors = min_t(unsigned, sectors, k.k->size - offset_into_extent); + if (bkey_start_offset(&cur.kbuf.k->k) == end) + break; - bch2_cut_front(POS(k.k->p.inode, - bkey_start_offset(k.k) + - offset_into_extent), - cur.k); - bch2_key_resize(&cur.k->k, sectors); - cur.k->k.p = iter.pos; - cur.k->k.p.offset += cur.k->k.size; + start = cur.kbuf.k->k.p.offset; - if (have_extent) { + if (!bkey_deleted(&prev.kbuf.k->k)) { bch2_trans_unlock(trans); - ret = bch2_fill_extent(c, info, - bkey_i_to_s_c(prev.k), 0); + ret = bch2_fill_extent(c, info, &prev); if (ret) - break; + goto err; } - bkey_copy(prev.k, cur.k); - have_extent = true; - - bch2_btree_iter_set_pos(trans, &iter, - POS(iter.pos.inode, iter.pos.offset + sectors)); + bch2_bkey_buf_copy(&prev.kbuf, c, cur.kbuf.k); + prev.flags = cur.flags; } - bch2_trans_iter_exit(trans, &iter); - if (!ret && have_extent) { + if (!bkey_deleted(&prev.kbuf.k->k)) { bch2_trans_unlock(trans); - ret = bch2_fill_extent(c, info, bkey_i_to_s_c(prev.k), - FIEMAP_EXTENT_LAST); + prev.flags |= FIEMAP_EXTENT_LAST; + ret = bch2_fill_extent(c, info, &prev); } - +err: bch2_trans_put(trans); - bch2_bkey_buf_exit(&cur, c); - bch2_bkey_buf_exit(&prev, c); - return ret < 0 ? ret : 0; + bch2_bkey_buf_exit(&cur.kbuf, c); + bch2_bkey_buf_exit(&prev.kbuf, c); + + return bch2_err_class(ret < 0 ? ret : 0); } static const struct vm_operations_struct bch_vm_ops = { @@ -1449,6 +1599,165 @@ static int bch2_open(struct inode *vinode, struct file *file) return generic_file_open(vinode, file); } +/* bcachefs inode flags -> FS_IOC_GETFLAGS: */ +static const __maybe_unused unsigned bch_flags_to_uflags[] = { + [__BCH_INODE_sync] = FS_SYNC_FL, + [__BCH_INODE_immutable] = FS_IMMUTABLE_FL, + [__BCH_INODE_append] = FS_APPEND_FL, + [__BCH_INODE_nodump] = FS_NODUMP_FL, + [__BCH_INODE_noatime] = FS_NOATIME_FL, +}; + +/* bcachefs inode flags -> FS_IOC_FSGETXATTR: */ +static const __maybe_unused unsigned bch_flags_to_xflags[] = { + [__BCH_INODE_sync] = FS_XFLAG_SYNC, + [__BCH_INODE_immutable] = FS_XFLAG_IMMUTABLE, + [__BCH_INODE_append] = FS_XFLAG_APPEND, + [__BCH_INODE_nodump] = FS_XFLAG_NODUMP, + [__BCH_INODE_noatime] = FS_XFLAG_NOATIME, +}; + +static int bch2_fileattr_get(struct dentry *dentry, + struct fileattr *fa) +{ + struct bch_inode_info *inode = to_bch_ei(d_inode(dentry)); + struct bch_fs *c = inode->v.i_sb->s_fs_info; + + fileattr_fill_xflags(fa, map_flags(bch_flags_to_xflags, inode->ei_inode.bi_flags)); + + if (inode->ei_inode.bi_fields_set & (1 << Inode_opt_project)) + fa->fsx_xflags |= FS_XFLAG_PROJINHERIT; + + if (bch2_inode_casefold(c, &inode->ei_inode)) + fa->flags |= FS_CASEFOLD_FL; + + fa->fsx_projid = inode->ei_qid.q[QTYP_PRJ]; + return 0; +} + +struct flags_set { + unsigned mask; + unsigned flags; + unsigned projid; + bool set_project; + bool set_casefold; + bool casefold; +}; + +static int fssetxattr_inode_update_fn(struct btree_trans *trans, + struct bch_inode_info *inode, + struct bch_inode_unpacked *bi, + void *p) +{ + struct bch_fs *c = trans->c; + struct flags_set *s = p; + + /* + * We're relying on btree locking here for exclusion with other ioctl + * calls - use the flags in the btree (@bi), not inode->i_flags: + */ + if (!S_ISREG(bi->bi_mode) && + !S_ISDIR(bi->bi_mode) && + (s->flags & (BCH_INODE_nodump|BCH_INODE_noatime)) != s->flags) + return -EINVAL; + + if (s->casefold != bch2_inode_casefold(c, bi)) { +#ifdef CONFIG_UNICODE + int ret = 0; + /* Not supported on individual files. */ + if (!S_ISDIR(bi->bi_mode)) + return -EOPNOTSUPP; + + /* + * Make sure the dir is empty, as otherwise we'd need to + * rehash everything and update the dirent keys. + */ + ret = bch2_empty_dir_trans(trans, inode_inum(inode)); + if (ret < 0) + return ret; + + ret = bch2_request_incompat_feature(c, bcachefs_metadata_version_casefolding); + if (ret) + return ret; + + bch2_check_set_feature(c, BCH_FEATURE_casefolding); + + bi->bi_casefold = s->casefold + 1; + bi->bi_fields_set |= BIT(Inode_opt_casefold); + +#else + printk(KERN_ERR "Cannot use casefolding on a kernel without CONFIG_UNICODE\n"); + return -EOPNOTSUPP; +#endif + } + + if (s->set_project) { + bi->bi_project = s->projid; + bi->bi_fields_set |= BIT(Inode_opt_project); + } + + bi->bi_flags &= ~s->mask; + bi->bi_flags |= s->flags; + + bi->bi_ctime = timespec_to_bch2_time(c, current_time(&inode->v)); + return 0; +} + +static int bch2_fileattr_set(struct mnt_idmap *idmap, + struct dentry *dentry, + struct fileattr *fa) +{ + struct bch_inode_info *inode = to_bch_ei(d_inode(dentry)); + struct bch_fs *c = inode->v.i_sb->s_fs_info; + struct flags_set s = {}; + int ret; + + if (fa->fsx_valid) { + fa->fsx_xflags &= ~FS_XFLAG_PROJINHERIT; + + s.mask = map_defined(bch_flags_to_xflags); + s.flags |= map_flags_rev(bch_flags_to_xflags, fa->fsx_xflags); + if (fa->fsx_xflags) + return -EOPNOTSUPP; + + if (fa->fsx_projid >= U32_MAX) + return -EINVAL; + + /* + * inode fields accessible via the xattr interface are stored with a +1 + * bias, so that 0 means unset: + */ + if ((inode->ei_inode.bi_project || + fa->fsx_projid) && + inode->ei_inode.bi_project != fa->fsx_projid + 1) { + s.projid = fa->fsx_projid + 1; + s.set_project = true; + } + } + + if (fa->flags_valid) { + s.mask = map_defined(bch_flags_to_uflags); + + s.set_casefold = true; + s.casefold = (fa->flags & FS_CASEFOLD_FL) != 0; + fa->flags &= ~FS_CASEFOLD_FL; + + s.flags |= map_flags_rev(bch_flags_to_uflags, fa->flags); + if (fa->flags) + return -EOPNOTSUPP; + } + + mutex_lock(&inode->ei_update_lock); + ret = bch2_subvol_is_ro(c, inode->ei_inum.subvol) ?: + (s.set_project + ? bch2_set_projid(c, inode, fa->fsx_projid) + : 0) ?: + bch2_write_inode(c, inode, fssetxattr_inode_update_fn, &s, + ATTR_CTIME); + mutex_unlock(&inode->ei_update_lock); + return ret; +} + static const struct file_operations bch_file_operations = { .open = bch2_open, .llseek = bch2_llseek, @@ -1476,6 +1785,8 @@ static const struct inode_operations bch_file_inode_operations = { .get_inode_acl = bch2_get_acl, .set_acl = bch2_set_acl, #endif + .fileattr_get = bch2_fileattr_get, + .fileattr_set = bch2_fileattr_set, }; static const struct inode_operations bch_dir_inode_operations = { @@ -1496,6 +1807,8 @@ static const struct inode_operations bch_dir_inode_operations = { .get_inode_acl = bch2_get_acl, .set_acl = bch2_set_acl, #endif + .fileattr_get = bch2_fileattr_get, + .fileattr_set = bch2_fileattr_set, }; static const struct file_operations bch_dir_file_operations = { @@ -1518,6 +1831,8 @@ static const struct inode_operations bch_symlink_inode_operations = { .get_inode_acl = bch2_get_acl, .set_acl = bch2_set_acl, #endif + .fileattr_get = bch2_fileattr_get, + .fileattr_set = bch2_fileattr_set, }; static const struct inode_operations bch_special_inode_operations = { @@ -1528,6 +1843,8 @@ static const struct inode_operations bch_special_inode_operations = { .get_inode_acl = bch2_get_acl, .set_acl = bch2_set_acl, #endif + .fileattr_get = bch2_fileattr_get, + .fileattr_set = bch2_fileattr_set, }; static const struct address_space_operations bch_address_space_operations = { @@ -2185,10 +2502,9 @@ static int bch2_fs_get_tree(struct fs_context *fc) bch2_opts_apply(&c->opts, opts); - /* - * need to initialise sb and set c->vfs_sb _before_ starting fs, - * for blk_holder_ops - */ + ret = bch2_fs_start(c); + if (ret) + goto err_stop_fs; sb = sget(fc->fs_type, NULL, bch2_set_super, fc->sb_flags|SB_NOSEC, c); ret = PTR_ERR_OR_ZERO(sb); @@ -2250,9 +2566,10 @@ got_sb: sb->s_shrink->seeks = 0; - ret = bch2_fs_start(c); - if (ret) - goto err_put_super; +#ifdef CONFIG_UNICODE + sb->s_encoding = c->cf_encoding; +#endif + generic_set_sb_d_ops(sb); vinode = bch2_vfs_inode_get(c, BCACHEFS_ROOT_SUBVOL_INUM); ret = PTR_ERR_OR_ZERO(vinode); diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c index 18308f3d64a1..7b25cedd3e40 100644 --- a/fs/bcachefs/fsck.c +++ b/fs/bcachefs/fsck.c @@ -321,6 +321,31 @@ static inline bool inode_should_reattach(struct bch_inode_unpacked *inode) inode->bi_subvol == BCACHEFS_ROOT_SUBVOL) return false; + /* + * Subvolume roots are special: older versions of subvolume roots may be + * disconnected, it's only the newest version that matters. + * + * We only keep a single dirent pointing to a subvolume root, i.e. + * older versions of snapshots will not have a different dirent pointing + * to the same subvolume root. + * + * This is because dirents that point to subvolumes are only visible in + * the parent subvolume - versioning is not needed - and keeping them + * around would break fsck, because when we're crossing subvolumes we + * don't have a consistent snapshot ID to do check the inode <-> dirent + * relationships. + * + * Thus, a subvolume root that's been renamed after a snapshot will have + * a disconnected older version - that's expected. + * + * Note that taking a snapshot always updates the root inode (to update + * the dirent backpointer), so a subvolume root inode with + * BCH_INODE_has_child_snapshot is never visible. + */ + if (inode->bi_subvol && + (inode->bi_flags & BCH_INODE_has_child_snapshot)) + return false; + return !inode->bi_dir && !(inode->bi_flags & BCH_INODE_unlinked); } @@ -1007,6 +1032,23 @@ static int check_inode_dirent_inode(struct btree_trans *trans, if (ret && !bch2_err_matches(ret, ENOENT)) return ret; + if ((ret || dirent_points_to_inode_nowarn(d, inode)) && + inode->bi_subvol && + (inode->bi_flags & BCH_INODE_has_child_snapshot)) { + /* Older version of a renamed subvolume root: we won't have a + * correct dirent for it. That's expected, see + * inode_should_reattach(). + * + * We don't clear the backpointer field when doing the rename + * because there might be arbitrarily many versions in older + * snapshots. + */ + inode->bi_dir = 0; + inode->bi_dir_offset = 0; + *write_inode = true; + goto out; + } + if (fsck_err_on(ret, trans, inode_points_to_missing_dirent, "inode points to missing dirent\n%s", @@ -1027,7 +1069,7 @@ static int check_inode_dirent_inode(struct btree_trans *trans, inode->bi_dir_offset = 0; *write_inode = true; } - +out: ret = 0; fsck_err: bch2_trans_iter_exit(trans, &dirent_iter); diff --git a/fs/bcachefs/inode.h b/fs/bcachefs/inode.h index f82cfbf460d0..c74af15b14f2 100644 --- a/fs/bcachefs/inode.h +++ b/fs/bcachefs/inode.h @@ -243,6 +243,14 @@ static inline unsigned bkey_inode_mode(struct bkey_s_c k) } } +static inline bool bch2_inode_casefold(struct bch_fs *c, const struct bch_inode_unpacked *bi) +{ + /* inode apts are stored with a +1 bias: 0 means "unset, use fs opt" */ + return bi->bi_casefold + ? bi->bi_casefold - 1 + : c->opts.casefold; +} + /* i_nlink: */ static inline unsigned nlink_bias(umode_t mode) diff --git a/fs/bcachefs/inode_format.h b/fs/bcachefs/inode_format.h index 117110af1e3f..87e193e8ed25 100644 --- a/fs/bcachefs/inode_format.h +++ b/fs/bcachefs/inode_format.h @@ -103,7 +103,8 @@ struct bch_inode_generation { x(bi_parent_subvol, 32) \ x(bi_nocow, 8) \ x(bi_depth, 32) \ - x(bi_inodes_32bit, 8) + x(bi_inodes_32bit, 8) \ + x(bi_casefold, 8) /* subset of BCH_INODE_FIELDS */ #define BCH_INODE_OPTS() \ @@ -117,7 +118,8 @@ struct bch_inode_generation { x(background_target, 16) \ x(erasure_code, 16) \ x(nocow, 8) \ - x(inodes_32bit, 8) + x(inodes_32bit, 8) \ + x(casefold, 8) enum inode_opt_id { #define x(name, ...) \ @@ -137,8 +139,7 @@ enum inode_opt_id { x(i_sectors_dirty, 6) \ x(unlinked, 7) \ x(backptr_untrusted, 8) \ - x(has_child_snapshot, 9) \ - x(casefolded, 10) + x(has_child_snapshot, 9) /* bits 20+ reserved for packed fields below: */ diff --git a/fs/bcachefs/io_read.c b/fs/bcachefs/io_read.c index 417bb0c7bbfa..def4a26a3b45 100644 --- a/fs/bcachefs/io_read.c +++ b/fs/bcachefs/io_read.c @@ -487,6 +487,8 @@ static void bch2_rbio_retry(struct work_struct *work) .inum = rbio->read_pos.inode, }; struct bch_io_failures failed = { .nr = 0 }; + int orig_error = rbio->ret; + struct btree_trans *trans = bch2_trans_get(c); trace_io_read_retry(&rbio->bio); @@ -519,7 +521,9 @@ static void bch2_rbio_retry(struct work_struct *work) if (ret) { rbio->ret = ret; rbio->bio.bi_status = BLK_STS_IOERR; - } else { + } else if (orig_error != -BCH_ERR_data_read_retry_csum_err_maybe_userspace && + orig_error != -BCH_ERR_data_read_ptr_stale_race && + !failed.nr) { struct printbuf buf = PRINTBUF; lockrestart_do(trans, @@ -977,7 +981,8 @@ retry_pick: goto err; } - if (unlikely(bch2_csum_type_is_encryption(pick.crc.csum_type)) && !c->chacha20) { + if (unlikely(bch2_csum_type_is_encryption(pick.crc.csum_type)) && + !c->chacha20_key_set) { struct printbuf buf = PRINTBUF; bch2_read_err_msg_trans(trans, &buf, orig, read_pos); prt_printf(&buf, "attempting to read encrypted data without encryption key\n "); @@ -1344,14 +1349,16 @@ err: bch2_trans_iter_exit(trans, &iter); - if (ret) { - struct printbuf buf = PRINTBUF; - lockrestart_do(trans, - bch2_inum_offset_err_msg_trans(trans, &buf, inum, - bvec_iter.bi_sector << 9)); - prt_printf(&buf, "read error: %s", bch2_err_str(ret)); - bch_err_ratelimited(c, "%s", buf.buf); - printbuf_exit(&buf); + if (unlikely(ret)) { + if (ret != -BCH_ERR_extent_poisoned) { + struct printbuf buf = PRINTBUF; + lockrestart_do(trans, + bch2_inum_offset_err_msg_trans(trans, &buf, inum, + bvec_iter.bi_sector << 9)); + prt_printf(&buf, "data read error: %s", bch2_err_str(ret)); + bch_err_ratelimited(c, "%s", buf.buf); + printbuf_exit(&buf); + } rbio->bio.bi_status = BLK_STS_IOERR; rbio->ret = ret; diff --git a/fs/bcachefs/io_write.c b/fs/bcachefs/io_write.c index a418fa62f09d..c1237da079ed 100644 --- a/fs/bcachefs/io_write.c +++ b/fs/bcachefs/io_write.c @@ -255,6 +255,27 @@ static inline int bch2_extent_update_i_size_sectors(struct btree_trans *trans, } if (i_sectors_delta) { + s64 bi_sectors = le64_to_cpu(inode->v.bi_sectors); + if (unlikely(bi_sectors + i_sectors_delta < 0)) { + struct bch_fs *c = trans->c; + struct printbuf buf = PRINTBUF; + bch2_log_msg_start(c, &buf); + prt_printf(&buf, "inode %llu i_sectors underflow: %lli + %lli < 0", + extent_iter->pos.inode, bi_sectors, i_sectors_delta); + + bool repeat = false, print = false, suppress = false; + bch2_count_fsck_err(c, inode_i_sectors_underflow, buf.buf, + &repeat, &print, &suppress); + if (print) + bch2_print_str(c, buf.buf); + printbuf_exit(&buf); + + if (i_sectors_delta < 0) + i_sectors_delta = -bi_sectors; + else + i_sectors_delta = 0; + } + le64_add_cpu(&inode->v.bi_sectors, i_sectors_delta); inode_update_flags = 0; } diff --git a/fs/bcachefs/journal.c b/fs/bcachefs/journal.c index d8f74b6d0a75..bb45d3634194 100644 --- a/fs/bcachefs/journal.c +++ b/fs/bcachefs/journal.c @@ -281,7 +281,24 @@ static void __journal_entry_close(struct journal *j, unsigned closed_val, bool t sectors = vstruct_blocks_plus(buf->data, c->block_bits, buf->u64s_reserved) << c->block_bits; - BUG_ON(sectors > buf->sectors); + if (unlikely(sectors > buf->sectors)) { + struct printbuf err = PRINTBUF; + err.atomic++; + + prt_printf(&err, "journal entry overran reserved space: %u > %u\n", + sectors, buf->sectors); + prt_printf(&err, "buf u64s %u u64s reserved %u cur_entry_u64s %u block_bits %u\n", + le32_to_cpu(buf->data->u64s), buf->u64s_reserved, + j->cur_entry_u64s, + c->block_bits); + prt_printf(&err, "fatal error - emergency read only"); + bch2_journal_halt_locked(j); + + bch_err(c, "%s", err.buf); + printbuf_exit(&err); + return; + } + buf->sectors = sectors; /* @@ -1462,8 +1479,6 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq) j->last_empty_seq = cur_seq - 1; /* to match j->seq */ spin_lock(&j->lock); - - set_bit(JOURNAL_running, &j->flags); j->last_flush_write = jiffies; j->reservations.idx = journal_cur_seq(j); @@ -1474,6 +1489,21 @@ int bch2_fs_journal_start(struct journal *j, u64 cur_seq) return 0; } +void bch2_journal_set_replay_done(struct journal *j) +{ + /* + * journal_space_available must happen before setting JOURNAL_running + * JOURNAL_running must happen before JOURNAL_replay_done + */ + spin_lock(&j->lock); + bch2_journal_space_available(j); + + set_bit(JOURNAL_need_flush_write, &j->flags); + set_bit(JOURNAL_running, &j->flags); + set_bit(JOURNAL_replay_done, &j->flags); + spin_unlock(&j->lock); +} + /* init/exit: */ void bch2_dev_journal_exit(struct bch_dev *ca) diff --git a/fs/bcachefs/journal.h b/fs/bcachefs/journal.h index 47828771f9c2..641e20c05a14 100644 --- a/fs/bcachefs/journal.h +++ b/fs/bcachefs/journal.h @@ -437,12 +437,6 @@ static inline int bch2_journal_error(struct journal *j) struct bch_dev; -static inline void bch2_journal_set_replay_done(struct journal *j) -{ - BUG_ON(!test_bit(JOURNAL_running, &j->flags)); - set_bit(JOURNAL_replay_done, &j->flags); -} - void bch2_journal_unblock(struct journal *); void bch2_journal_block(struct journal *); struct journal_buf *bch2_next_write_buffer_flush_journal_buf(struct journal *, u64, bool *); @@ -459,6 +453,7 @@ void bch2_dev_journal_stop(struct journal *, struct bch_dev *); void bch2_fs_journal_stop(struct journal *); int bch2_fs_journal_start(struct journal *, u64); +void bch2_journal_set_replay_done(struct journal *); void bch2_dev_journal_exit(struct bch_dev *); int bch2_dev_journal_init(struct bch_dev *, struct bch_sb *); diff --git a/fs/bcachefs/journal_io.c b/fs/bcachefs/journal_io.c index 1b7961f4f609..ded18a94ed02 100644 --- a/fs/bcachefs/journal_io.c +++ b/fs/bcachefs/journal_io.c @@ -19,6 +19,7 @@ #include <linux/ioprio.h> #include <linux/string_choices.h> +#include <linux/sched/sysctl.h> void bch2_journal_pos_from_member_info_set(struct bch_fs *c) { @@ -1262,7 +1263,8 @@ int bch2_journal_read(struct bch_fs *c, degraded = true; } - closure_sync(&jlist.cl); + while (closure_sync_timeout(&jlist.cl, sysctl_hung_task_timeout_secs * HZ / 2)) + ; if (jlist.ret) return jlist.ret; @@ -1460,7 +1462,7 @@ fsck_err: static void journal_advance_devs_to_next_bucket(struct journal *j, struct dev_alloc_list *devs, - unsigned sectors, u64 seq) + unsigned sectors, __le64 seq) { struct bch_fs *c = container_of(j, struct bch_fs, journal); @@ -1782,7 +1784,7 @@ static CLOSURE_CALLBACK(journal_write_submit) struct bch_dev *ca = bch2_dev_get_ioref(c, ptr->dev, WRITE); if (!ca) { /* XXX: fix this */ - bch_err(c, "missing device for journal write\n"); + bch_err(c, "missing device %u for journal write", ptr->dev); continue; } diff --git a/fs/bcachefs/journal_reclaim.c b/fs/bcachefs/journal_reclaim.c index 5d1547aa118a..976464d8a695 100644 --- a/fs/bcachefs/journal_reclaim.c +++ b/fs/bcachefs/journal_reclaim.c @@ -252,7 +252,10 @@ void bch2_journal_space_available(struct journal *j) bch2_journal_set_watermark(j); out: - j->cur_entry_sectors = !ret ? j->space[journal_space_discarded].next_entry : 0; + j->cur_entry_sectors = !ret + ? round_down(j->space[journal_space_discarded].next_entry, + block_sectors(c)) + : 0; j->cur_entry_error = ret; if (!ret) @@ -263,10 +266,11 @@ out: static bool should_discard_bucket(struct journal *j, struct journal_device *ja) { - bool ret; - spin_lock(&j->lock); - ret = ja->discard_idx != ja->dirty_idx_ondisk; + unsigned min_free = max(4, ja->nr / 8); + + bool ret = bch2_journal_dev_buckets_available(j, ja, journal_space_discarded) < min_free && + ja->discard_idx != ja->dirty_idx_ondisk; spin_unlock(&j->lock); return ret; diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c index fc396b9fa754..dfdbb9259985 100644 --- a/fs/bcachefs/move.c +++ b/fs/bcachefs/move.c @@ -784,7 +784,8 @@ static int __bch2_move_data_phys(struct moving_context *ctxt, goto err; ret = bch2_btree_write_buffer_tryflush(trans); - bch_err_msg(c, ret, "flushing btree write buffer"); + if (!bch2_err_matches(ret, EROFS)) + bch_err_msg(c, ret, "flushing btree write buffer"); if (ret) goto err; diff --git a/fs/bcachefs/movinggc.c b/fs/bcachefs/movinggc.c index 159410c50861..96873372b516 100644 --- a/fs/bcachefs/movinggc.c +++ b/fs/bcachefs/movinggc.c @@ -356,6 +356,13 @@ static int bch2_copygc_thread(void *arg) set_freezable(); + /* + * Data move operations can't run until after check_snapshots has + * completed, and bch2_snapshot_is_ancestor() is available. + */ + kthread_wait_freezable(c->recovery_pass_done > BCH_RECOVERY_PASS_check_snapshots || + kthread_should_stop()); + bch2_move_stats_init(&move_stats, "copygc"); bch2_moving_ctxt_init(&ctxt, c, NULL, &move_stats, writepoint_ptr(&c->copygc_write_point), diff --git a/fs/bcachefs/movinggc.h b/fs/bcachefs/movinggc.h index ea181fef5bc9..d1885cf67a45 100644 --- a/fs/bcachefs/movinggc.h +++ b/fs/bcachefs/movinggc.h @@ -5,6 +5,15 @@ unsigned long bch2_copygc_wait_amount(struct bch_fs *); void bch2_copygc_wait_to_text(struct printbuf *, struct bch_fs *); +static inline void bch2_copygc_wakeup(struct bch_fs *c) +{ + rcu_read_lock(); + struct task_struct *p = rcu_dereference(c->copygc_thread); + if (p) + wake_up_process(p); + rcu_read_unlock(); +} + void bch2_copygc_stop(struct bch_fs *); int bch2_copygc_start(struct bch_fs *); void bch2_fs_copygc_init(struct bch_fs *); diff --git a/fs/bcachefs/namei.c b/fs/bcachefs/namei.c index 0d65ea96f7a2..52c58c6d53d2 100644 --- a/fs/bcachefs/namei.c +++ b/fs/bcachefs/namei.c @@ -47,10 +47,6 @@ int bch2_create_trans(struct btree_trans *trans, if (ret) goto err; - /* Inherit casefold state from parent. */ - if (S_ISDIR(mode)) - new_inode->bi_flags |= dir_u->bi_flags & BCH_INODE_casefolded; - if (!(flags & BCH_CREATE_SNAPSHOT)) { /* Normal create path - allocate a new inode: */ bch2_inode_init_late(new_inode, now, uid, gid, mode, rdev, dir_u); @@ -347,6 +343,9 @@ bool bch2_reinherit_attrs(struct bch_inode_unpacked *dst_u, bool ret = false; for (id = 0; id < Inode_opt_nr; id++) { + if (!S_ISDIR(dst_u->bi_mode) && id == Inode_opt_casefold) + continue; + /* Skip attributes that were explicitly set on this inode */ if (dst_u->bi_fields_set & (1 << id)) continue; diff --git a/fs/bcachefs/opts.h b/fs/bcachefs/opts.h index 4d06313076ff..dfb14810124c 100644 --- a/fs/bcachefs/opts.h +++ b/fs/bcachefs/opts.h @@ -228,6 +228,11 @@ enum fsck_err_opts { OPT_BOOL(), \ BCH_SB_ERASURE_CODE, false, \ NULL, "Enable erasure coding (DO NOT USE YET)") \ + x(casefold, u8, \ + OPT_FS|OPT_INODE|OPT_FORMAT, \ + OPT_BOOL(), \ + BCH_SB_CASEFOLD, false, \ + NULL, "Dirent lookups are casefolded") \ x(inodes_32bit, u8, \ OPT_FS|OPT_INODE|OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME, \ OPT_BOOL(), \ diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c index c63fa53f30d2..4ccdfc1f34aa 100644 --- a/fs/bcachefs/rebalance.c +++ b/fs/bcachefs/rebalance.c @@ -262,7 +262,7 @@ int bch2_set_rebalance_needs_scan(struct bch_fs *c, u64 inum) int ret = bch2_trans_commit_do(c, NULL, NULL, BCH_TRANS_COMMIT_no_enospc, bch2_set_rebalance_needs_scan_trans(trans, inum)); - rebalance_wakeup(c); + bch2_rebalance_wakeup(c); return ret; } @@ -581,6 +581,13 @@ static int bch2_rebalance_thread(void *arg) set_freezable(); + /* + * Data move operations can't run until after check_snapshots has + * completed, and bch2_snapshot_is_ancestor() is available. + */ + kthread_wait_freezable(c->recovery_pass_done > BCH_RECOVERY_PASS_check_snapshots || + kthread_should_stop()); + bch2_moving_ctxt_init(&ctxt, c, NULL, &r->work_stats, writepoint_ptr(&c->rebalance_write_point), true); @@ -664,7 +671,7 @@ void bch2_rebalance_stop(struct bch_fs *c) c->rebalance.thread = NULL; if (p) { - /* for sychronizing with rebalance_wakeup() */ + /* for sychronizing with bch2_rebalance_wakeup() */ synchronize_rcu(); kthread_stop(p); diff --git a/fs/bcachefs/rebalance.h b/fs/bcachefs/rebalance.h index 62a3859d3823..e5e8eb4a2dd1 100644 --- a/fs/bcachefs/rebalance.h +++ b/fs/bcachefs/rebalance.h @@ -37,7 +37,7 @@ int bch2_set_rebalance_needs_scan_trans(struct btree_trans *, u64); int bch2_set_rebalance_needs_scan(struct bch_fs *, u64 inum); int bch2_set_fs_needs_rebalance(struct bch_fs *); -static inline void rebalance_wakeup(struct bch_fs *c) +static inline void bch2_rebalance_wakeup(struct bch_fs *c) { struct task_struct *p; diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c index 79fd18a5a07c..d6c4ef819d40 100644 --- a/fs/bcachefs/recovery.c +++ b/fs/bcachefs/recovery.c @@ -18,6 +18,7 @@ #include "journal_seq_blacklist.h" #include "logged_ops.h" #include "move.h" +#include "movinggc.h" #include "namei.h" #include "quota.h" #include "rebalance.h" @@ -389,9 +390,9 @@ int bch2_journal_replay(struct bch_fs *c) * Now, replay any remaining keys in the order in which they appear in * the journal, unpinning those journal entries as we go: */ - sort(keys_sorted.data, keys_sorted.nr, - sizeof(keys_sorted.data[0]), - journal_sort_seq_cmp, NULL); + sort_nonatomic(keys_sorted.data, keys_sorted.nr, + sizeof(keys_sorted.data[0]), + journal_sort_seq_cmp, NULL); darray_for_each(keys_sorted, kp) { cond_resched(); @@ -1125,14 +1126,17 @@ int bch2_fs_initialize(struct bch_fs *c) * journal_res_get() will crash if called before this has * set up the journal.pin FIFO and journal.cur pointer: */ - bch2_fs_journal_start(&c->journal, 1); - set_bit(BCH_FS_accounting_replay_done, &c->flags); - bch2_journal_set_replay_done(&c->journal); + ret = bch2_fs_journal_start(&c->journal, 1); + if (ret) + goto err; ret = bch2_fs_read_write_early(c); if (ret) goto err; + set_bit(BCH_FS_accounting_replay_done, &c->flags); + bch2_journal_set_replay_done(&c->journal); + for_each_member_device(c, ca) { ret = bch2_dev_usage_init(ca, false); if (ret) { @@ -1191,6 +1195,9 @@ int bch2_fs_initialize(struct bch_fs *c) c->recovery_pass_done = BCH_RECOVERY_PASS_NR - 1; + bch2_copygc_wakeup(c); + bch2_rebalance_wakeup(c); + if (enabled_qtypes(c)) { ret = bch2_fs_quota_read(c); if (ret) diff --git a/fs/bcachefs/recovery_passes.c b/fs/bcachefs/recovery_passes.c index 593ff142530d..22f72bb5b853 100644 --- a/fs/bcachefs/recovery_passes.c +++ b/fs/bcachefs/recovery_passes.c @@ -12,6 +12,7 @@ #include "journal.h" #include "lru.h" #include "logged_ops.h" +#include "movinggc.h" #include "rebalance.h" #include "recovery.h" #include "recovery_passes.h" @@ -262,49 +263,52 @@ int bch2_run_recovery_passes(struct bch_fs *c) */ c->opts.recovery_passes_exclude &= ~BCH_RECOVERY_PASS_set_may_go_rw; - while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns) && !ret) { - c->next_recovery_pass = c->curr_recovery_pass + 1; + spin_lock_irq(&c->recovery_pass_lock); - spin_lock_irq(&c->recovery_pass_lock); + while (c->curr_recovery_pass < ARRAY_SIZE(recovery_pass_fns) && !ret) { + unsigned prev_done = c->recovery_pass_done; unsigned pass = c->curr_recovery_pass; + c->next_recovery_pass = pass + 1; + if (c->opts.recovery_pass_last && - c->curr_recovery_pass > c->opts.recovery_pass_last) { - spin_unlock_irq(&c->recovery_pass_lock); + c->curr_recovery_pass > c->opts.recovery_pass_last) break; - } - if (!should_run_recovery_pass(c, pass)) { - c->curr_recovery_pass++; - c->recovery_pass_done = max(c->recovery_pass_done, pass); + if (should_run_recovery_pass(c, pass)) { spin_unlock_irq(&c->recovery_pass_lock); - continue; - } - spin_unlock_irq(&c->recovery_pass_lock); - - ret = bch2_run_recovery_pass(c, pass) ?: - bch2_journal_flush(&c->journal); - - if (!ret && !test_bit(BCH_FS_error, &c->flags)) - bch2_clear_recovery_pass_required(c, pass); - - spin_lock_irq(&c->recovery_pass_lock); - if (c->next_recovery_pass < c->curr_recovery_pass) { - /* - * bch2_run_explicit_recovery_pass() was called: we - * can't always catch -BCH_ERR_restart_recovery because - * it may have been called from another thread (btree - * node read completion) - */ - ret = 0; - c->recovery_passes_complete &= ~(~0ULL << c->curr_recovery_pass); - } else { - c->recovery_passes_complete |= BIT_ULL(pass); - c->recovery_pass_done = max(c->recovery_pass_done, pass); + ret = bch2_run_recovery_pass(c, pass) ?: + bch2_journal_flush(&c->journal); + + if (!ret && !test_bit(BCH_FS_error, &c->flags)) + bch2_clear_recovery_pass_required(c, pass); + spin_lock_irq(&c->recovery_pass_lock); + + if (c->next_recovery_pass < c->curr_recovery_pass) { + /* + * bch2_run_explicit_recovery_pass() was called: we + * can't always catch -BCH_ERR_restart_recovery because + * it may have been called from another thread (btree + * node read completion) + */ + ret = 0; + c->recovery_passes_complete &= ~(~0ULL << c->curr_recovery_pass); + } else { + c->recovery_passes_complete |= BIT_ULL(pass); + c->recovery_pass_done = max(c->recovery_pass_done, pass); + } } + c->curr_recovery_pass = c->next_recovery_pass; - spin_unlock_irq(&c->recovery_pass_lock); + + if (prev_done <= BCH_RECOVERY_PASS_check_snapshots && + c->recovery_pass_done > BCH_RECOVERY_PASS_check_snapshots) { + bch2_copygc_wakeup(c); + bch2_rebalance_wakeup(c); + } } + spin_unlock_irq(&c->recovery_pass_lock); + return ret; } diff --git a/fs/bcachefs/sb-downgrade.c b/fs/bcachefs/sb-downgrade.c index acb5d845841e..badd0e17ada5 100644 --- a/fs/bcachefs/sb-downgrade.c +++ b/fs/bcachefs/sb-downgrade.c @@ -20,6 +20,10 @@ * x(version, recovery_passes, errors...) */ #define UPGRADE_TABLE() \ + x(snapshot_2, \ + RECOVERY_PASS_ALL_FSCK, \ + BCH_FSCK_ERR_subvol_root_wrong_bi_subvol, \ + BCH_FSCK_ERR_subvol_not_master_and_not_snapshot) \ x(backpointers, \ RECOVERY_PASS_ALL_FSCK) \ x(inode_v3, \ diff --git a/fs/bcachefs/sb-errors_format.h b/fs/bcachefs/sb-errors_format.h index 5d43e3504386..3b69a924086f 100644 --- a/fs/bcachefs/sb-errors_format.h +++ b/fs/bcachefs/sb-errors_format.h @@ -46,7 +46,7 @@ enum bch_fsck_flags { x(btree_node_unsupported_version, 34, 0) \ x(btree_node_bset_older_than_sb_min, 35, 0) \ x(btree_node_bset_newer_than_sb, 36, 0) \ - x(btree_node_data_missing, 37, 0) \ + x(btree_node_data_missing, 37, FSCK_AUTOFIX) \ x(btree_node_bset_after_end, 38, 0) \ x(btree_node_replicas_sectors_written_mismatch, 39, 0) \ x(btree_node_replicas_data_mismatch, 40, 0) \ @@ -205,9 +205,9 @@ enum bch_fsck_flags { x(snapshot_bad_depth, 184, 0) \ x(snapshot_bad_skiplist, 185, 0) \ x(subvol_pos_bad, 186, 0) \ - x(subvol_not_master_and_not_snapshot, 187, 0) \ + x(subvol_not_master_and_not_snapshot, 187, FSCK_AUTOFIX) \ x(subvol_to_missing_root, 188, 0) \ - x(subvol_root_wrong_bi_subvol, 189, 0) \ + x(subvol_root_wrong_bi_subvol, 189, FSCK_AUTOFIX) \ x(bkey_in_missing_snapshot, 190, 0) \ x(inode_pos_inode_nonzero, 191, 0) \ x(inode_pos_blockdev_range, 192, 0) \ @@ -236,6 +236,9 @@ enum bch_fsck_flags { x(inode_has_child_snapshots_wrong, 287, 0) \ x(inode_unreachable, 210, FSCK_AUTOFIX) \ x(inode_journal_seq_in_future, 299, FSCK_AUTOFIX) \ + x(inode_i_sectors_underflow, 312, FSCK_AUTOFIX) \ + x(vfs_inode_i_blocks_underflow, 311, FSCK_AUTOFIX) \ + x(vfs_inode_i_blocks_not_zero_at_truncate, 313, FSCK_AUTOFIX) \ x(deleted_inode_but_clean, 211, FSCK_AUTOFIX) \ x(deleted_inode_missing, 212, FSCK_AUTOFIX) \ x(deleted_inode_is_dir, 213, FSCK_AUTOFIX) \ @@ -290,8 +293,8 @@ enum bch_fsck_flags { x(btree_node_bkey_bad_u64s, 260, 0) \ x(btree_node_topology_empty_interior_node, 261, 0) \ x(btree_ptr_v2_min_key_bad, 262, 0) \ - x(btree_root_unreadable_and_scan_found_nothing, 263, 0) \ - x(snapshot_node_missing, 264, 0) \ + x(btree_root_unreadable_and_scan_found_nothing, 263, FSCK_AUTOFIX) \ + x(snapshot_node_missing, 264, FSCK_AUTOFIX) \ x(dup_backpointer_to_bad_csum_extent, 265, 0) \ x(btree_bitmap_not_marked, 266, FSCK_AUTOFIX) \ x(sb_clean_entry_overrun, 267, 0) \ @@ -317,7 +320,9 @@ enum bch_fsck_flags { x(directory_size_mismatch, 303, FSCK_AUTOFIX) \ x(dirent_cf_name_too_big, 304, 0) \ x(dirent_stray_data_after_cf_name, 305, 0) \ - x(MAX, 308, 0) + x(rebalance_work_incorrectly_set, 309, FSCK_AUTOFIX) \ + x(rebalance_work_incorrectly_unset, 310, FSCK_AUTOFIX) \ + x(MAX, 314, 0) enum bch_sb_error_id { #define x(t, n, ...) BCH_FSCK_ERR_##t = n, diff --git a/fs/bcachefs/sb-members.c b/fs/bcachefs/sb-members.c index 116131f95815..72779912939b 100644 --- a/fs/bcachefs/sb-members.c +++ b/fs/bcachefs/sb-members.c @@ -15,9 +15,11 @@ void bch2_dev_missing(struct bch_fs *c, unsigned dev) bch2_fs_inconsistent(c, "pointer to nonexistent device %u", dev); } -void bch2_dev_bucket_missing(struct bch_fs *c, struct bpos bucket) +void bch2_dev_bucket_missing(struct bch_dev *ca, u64 bucket) { - bch2_fs_inconsistent(c, "pointer to nonexistent bucket %llu:%llu", bucket.inode, bucket.offset); + bch2_fs_inconsistent(ca->fs, + "pointer to nonexistent bucket %llu on device %s (valid range %u-%llu)", + bucket, ca->name, ca->mi.first_bucket, ca->mi.nbuckets); } #define x(t, n, ...) [n] = #t, diff --git a/fs/bcachefs/sb-members.h b/fs/bcachefs/sb-members.h index 06bb41a3f360..42786657522c 100644 --- a/fs/bcachefs/sb-members.h +++ b/fs/bcachefs/sb-members.h @@ -249,20 +249,23 @@ static inline struct bch_dev *bch2_dev_tryget(struct bch_fs *c, unsigned dev) static inline struct bch_dev *bch2_dev_bucket_tryget_noerror(struct bch_fs *c, struct bpos bucket) { struct bch_dev *ca = bch2_dev_tryget_noerror(c, bucket.inode); - if (ca && !bucket_valid(ca, bucket.offset)) { + if (ca && unlikely(!bucket_valid(ca, bucket.offset))) { bch2_dev_put(ca); ca = NULL; } return ca; } -void bch2_dev_bucket_missing(struct bch_fs *, struct bpos); +void bch2_dev_bucket_missing(struct bch_dev *, u64); static inline struct bch_dev *bch2_dev_bucket_tryget(struct bch_fs *c, struct bpos bucket) { - struct bch_dev *ca = bch2_dev_bucket_tryget_noerror(c, bucket); - if (!ca) - bch2_dev_bucket_missing(c, bucket); + struct bch_dev *ca = bch2_dev_tryget(c, bucket.inode); + if (ca && unlikely(!bucket_valid(ca, bucket.offset))) { + bch2_dev_bucket_missing(ca, bucket.offset); + bch2_dev_put(ca); + ca = NULL; + } return ca; } diff --git a/fs/bcachefs/snapshot.c b/fs/bcachefs/snapshot.c index b7de29aed839..fec569c7deb1 100644 --- a/fs/bcachefs/snapshot.c +++ b/fs/bcachefs/snapshot.c @@ -396,7 +396,7 @@ u32 bch2_snapshot_tree_oldest_subvol(struct bch_fs *c, u32 snapshot_root) u32 subvol = 0, s; rcu_read_lock(); - while (id) { + while (id && bch2_snapshot_exists(c, id)) { s = snapshot_t(c, id)->subvol; if (s && (!subvol || s < subvol)) diff --git a/fs/bcachefs/str_hash.h b/fs/bcachefs/str_hash.h index 09a354a26c3b..0c1a00539bd1 100644 --- a/fs/bcachefs/str_hash.h +++ b/fs/bcachefs/str_hash.h @@ -33,7 +33,7 @@ bch2_str_hash_opt_to_type(struct bch_fs *c, enum bch_str_hash_opts opt) struct bch_hash_info { u8 type; - struct unicode_map *cf_encoding; + struct unicode_map *cf_encoding; /* * For crc32 or crc64 string hashes the first key value of * the siphash_key (k0) is used as the key. @@ -44,11 +44,10 @@ struct bch_hash_info { static inline struct bch_hash_info bch2_hash_info_init(struct bch_fs *c, const struct bch_inode_unpacked *bi) { - /* XXX ick */ struct bch_hash_info info = { .type = INODE_STR_HASH(bi), #ifdef CONFIG_UNICODE - .cf_encoding = !!(bi->bi_flags & BCH_INODE_casefolded) ? c->cf_encoding : NULL, + .cf_encoding = bch2_inode_casefold(c, bi) ? c->cf_encoding : NULL, #endif .siphash_key = { .k0 = bi->bi_hash_seed } }; diff --git a/fs/bcachefs/subvolume.c b/fs/bcachefs/subvolume.c index 5537283d0bea..d0209f7658bb 100644 --- a/fs/bcachefs/subvolume.c +++ b/fs/bcachefs/subvolume.c @@ -6,6 +6,7 @@ #include "errcode.h" #include "error.h" #include "fs.h" +#include "recovery_passes.h" #include "snapshot.h" #include "subvolume.h" @@ -44,8 +45,8 @@ static int check_subvol(struct btree_trans *trans, ret = bch2_snapshot_lookup(trans, snapid, &snapshot); if (bch2_err_matches(ret, ENOENT)) - bch_err(c, "subvolume %llu points to nonexistent snapshot %u", - k.k->p.offset, snapid); + return bch2_run_explicit_recovery_pass(c, + BCH_RECOVERY_PASS_reconstruct_snapshots) ?: ret; if (ret) return ret; diff --git a/fs/bcachefs/super-io.c b/fs/bcachefs/super-io.c index e27422b6d9c6..cb5d960aed92 100644 --- a/fs/bcachefs/super-io.c +++ b/fs/bcachefs/super-io.c @@ -73,14 +73,30 @@ int bch2_set_version_incompat(struct bch_fs *c, enum bcachefs_metadata_version v ? 0 : -BCH_ERR_may_not_use_incompat_feature; + mutex_lock(&c->sb_lock); if (!ret) { - mutex_lock(&c->sb_lock); SET_BCH_SB_VERSION_INCOMPAT(c->disk_sb.sb, max(BCH_SB_VERSION_INCOMPAT(c->disk_sb.sb), version)); bch2_write_super(c); - mutex_unlock(&c->sb_lock); + } else { + darray_for_each(c->incompat_versions_requested, i) + if (version == *i) + goto out; + + darray_push(&c->incompat_versions_requested, version); + struct printbuf buf = PRINTBUF; + prt_str(&buf, "requested incompat feature "); + bch2_version_to_text(&buf, version); + prt_str(&buf, " currently not enabled"); + prt_printf(&buf, "\n set version_upgrade=incompat to enable"); + + bch_notice(c, "%s", buf.buf); + printbuf_exit(&buf); } +out: + mutex_unlock(&c->sb_lock); + return ret; } @@ -1086,7 +1102,8 @@ int bch2_write_super(struct bch_fs *c) prt_str(&buf, ")"); bch2_fs_fatal_error(c, ": %s", buf.buf); printbuf_exit(&buf); - return -BCH_ERR_sb_not_downgraded; + ret = -BCH_ERR_sb_not_downgraded; + goto out; } darray_for_each(online_devices, ca) { diff --git a/fs/bcachefs/super.c b/fs/bcachefs/super.c index a58edde43bee..84a37d971ffd 100644 --- a/fs/bcachefs/super.c +++ b/fs/bcachefs/super.c @@ -70,14 +70,10 @@ #include <linux/percpu.h> #include <linux/random.h> #include <linux/sysfs.h> -#include <crypto/hash.h> MODULE_LICENSE("GPL"); MODULE_AUTHOR("Kent Overstreet <kent.overstreet@gmail.com>"); MODULE_DESCRIPTION("bcachefs filesystem"); -MODULE_SOFTDEP("pre: chacha20"); -MODULE_SOFTDEP("pre: poly1305"); -MODULE_SOFTDEP("pre: xxhash"); const char * const bch2_fs_flag_strs[] = { #define x(n) #n, @@ -381,6 +377,11 @@ void bch2_fs_read_only(struct bch_fs *c) bch_verbose(c, "marking filesystem clean"); bch2_fs_mark_clean(c); } else { + /* Make sure error counts/counters are persisted */ + mutex_lock(&c->sb_lock); + bch2_write_super(c); + mutex_unlock(&c->sb_lock); + bch_verbose(c, "done going read-only, filesystem not clean"); } } @@ -422,32 +423,6 @@ bool bch2_fs_emergency_read_only_locked(struct bch_fs *c) return ret; } -static int bch2_fs_read_write_late(struct bch_fs *c) -{ - int ret; - - /* - * Data move operations can't run until after check_snapshots has - * completed, and bch2_snapshot_is_ancestor() is available. - * - * Ideally we'd start copygc/rebalance earlier instead of waiting for - * all of recovery/fsck to complete: - */ - ret = bch2_copygc_start(c); - if (ret) { - bch_err(c, "error starting copygc thread"); - return ret; - } - - ret = bch2_rebalance_start(c); - if (ret) { - bch_err(c, "error starting rebalance thread"); - return ret; - } - - return 0; -} - static int __bch2_fs_read_write(struct bch_fs *c, bool early) { int ret; @@ -470,29 +445,28 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early) clear_bit(BCH_FS_clean_shutdown, &c->flags); + __for_each_online_member(c, ca, BIT(BCH_MEMBER_STATE_rw), READ) { + bch2_dev_allocator_add(c, ca); + percpu_ref_reinit(&ca->io_ref[WRITE]); + } + bch2_recalc_capacity(c); + /* * First journal write must be a flush write: after a clean shutdown we * don't read the journal, so the first journal write may end up * overwriting whatever was there previously, and there must always be * at least one non-flush write in the journal or recovery will fail: */ + spin_lock(&c->journal.lock); set_bit(JOURNAL_need_flush_write, &c->journal.flags); set_bit(JOURNAL_running, &c->journal.flags); - - __for_each_online_member(c, ca, BIT(BCH_MEMBER_STATE_rw), READ) { - bch2_dev_allocator_add(c, ca); - percpu_ref_reinit(&ca->io_ref[WRITE]); - } - bch2_recalc_capacity(c); + bch2_journal_space_available(&c->journal); + spin_unlock(&c->journal.lock); ret = bch2_fs_mark_dirty(c); if (ret) goto err; - spin_lock(&c->journal.lock); - bch2_journal_space_available(&c->journal); - spin_unlock(&c->journal.lock); - ret = bch2_journal_reclaim_start(&c->journal); if (ret) goto err; @@ -508,10 +482,17 @@ static int __bch2_fs_read_write(struct bch_fs *c, bool early) atomic_long_inc(&c->writes[i]); } #endif - if (!early) { - ret = bch2_fs_read_write_late(c); - if (ret) - goto err; + + ret = bch2_copygc_start(c); + if (ret) { + bch_err_msg(c, ret, "error starting copygc thread"); + goto err; + } + + ret = bch2_rebalance_start(c); + if (ret) { + bch_err_msg(c, ret, "error starting rebalance thread"); + goto err; } bch2_do_discards(c); @@ -555,8 +536,13 @@ static void __bch2_fs_free(struct bch_fs *c) for (unsigned i = 0; i < BCH_TIME_STAT_NR; i++) bch2_time_stats_exit(&c->times[i]); +#ifdef CONFIG_UNICODE + utf8_unload(c->cf_encoding); +#endif + bch2_find_btree_nodes_exit(&c->found_btree_nodes); bch2_free_pending_node_rewrites(c); + bch2_free_fsck_errs(c); bch2_fs_accounting_exit(c); bch2_fs_sb_errors_exit(c); bch2_fs_counters_exit(c); @@ -593,6 +579,7 @@ static void __bch2_fs_free(struct bch_fs *c) free_percpu(c->online_reserved); } + darray_exit(&c->incompat_versions_requested); darray_exit(&c->btree_roots_extra); free_percpu(c->pcpu); free_percpu(c->usage); @@ -845,25 +832,6 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts) if (ret) goto err; -#ifdef CONFIG_UNICODE - /* Default encoding until we can potentially have more as an option. */ - c->cf_encoding = utf8_load(BCH_FS_DEFAULT_UTF8_ENCODING); - if (IS_ERR(c->cf_encoding)) { - printk(KERN_ERR "Cannot load UTF-8 encoding for filesystem. Version: %u.%u.%u", - unicode_major(BCH_FS_DEFAULT_UTF8_ENCODING), - unicode_minor(BCH_FS_DEFAULT_UTF8_ENCODING), - unicode_rev(BCH_FS_DEFAULT_UTF8_ENCODING)); - ret = -EINVAL; - goto err; - } -#else - if (c->sb.features & BIT_ULL(BCH_FEATURE_casefolding)) { - printk(KERN_ERR "Cannot mount a filesystem with casefolding on a kernel without CONFIG_UNICODE\n"); - ret = -EINVAL; - goto err; - } -#endif - pr_uuid(&name, c->sb.user_uuid.b); ret = name.allocation_failure ? -BCH_ERR_ENOMEM_fs_name_alloc : 0; if (ret) @@ -963,6 +931,29 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts) if (ret) goto err; +#ifdef CONFIG_UNICODE + /* Default encoding until we can potentially have more as an option. */ + c->cf_encoding = utf8_load(BCH_FS_DEFAULT_UTF8_ENCODING); + if (IS_ERR(c->cf_encoding)) { + printk(KERN_ERR "Cannot load UTF-8 encoding for filesystem. Version: %u.%u.%u", + unicode_major(BCH_FS_DEFAULT_UTF8_ENCODING), + unicode_minor(BCH_FS_DEFAULT_UTF8_ENCODING), + unicode_rev(BCH_FS_DEFAULT_UTF8_ENCODING)); + ret = -EINVAL; + goto err; + } + bch_info(c, "Using encoding defined by superblock: utf8-%u.%u.%u", + unicode_major(BCH_FS_DEFAULT_UTF8_ENCODING), + unicode_minor(BCH_FS_DEFAULT_UTF8_ENCODING), + unicode_rev(BCH_FS_DEFAULT_UTF8_ENCODING)); +#else + if (c->sb.features & BIT_ULL(BCH_FEATURE_casefolding)) { + printk(KERN_ERR "Cannot mount a filesystem with casefolding on a kernel without CONFIG_UNICODE\n"); + ret = -EINVAL; + goto err; + } +#endif + for (i = 0; i < c->sb.nr_devices; i++) { if (!bch2_member_exists(c->disk_sb.sb, i)) continue; @@ -1002,12 +993,6 @@ static void print_mount_opts(struct bch_fs *c) prt_str(&p, "starting version "); bch2_version_to_text(&p, c->sb.version); - if (c->opts.read_only) { - prt_str(&p, " opts="); - first = false; - prt_printf(&p, "ro"); - } - for (i = 0; i < bch2_opts_nr; i++) { const struct bch_option *opt = &bch2_opt_table[i]; u64 v = bch2_opt_get_by_id(&c->opts, i); @@ -1023,10 +1008,49 @@ static void print_mount_opts(struct bch_fs *c) bch2_opt_to_text(&p, c, c->disk_sb.sb, opt, v, OPT_SHOW_MOUNT_STYLE); } + if (c->sb.version_incompat_allowed != c->sb.version) { + prt_printf(&p, "\n allowing incompatible features above "); + bch2_version_to_text(&p, c->sb.version_incompat_allowed); + } + bch_info(c, "%s", p.buf); printbuf_exit(&p); } +static bool bch2_fs_may_start(struct bch_fs *c) +{ + struct bch_dev *ca; + unsigned i, flags = 0; + + if (c->opts.very_degraded) + flags |= BCH_FORCE_IF_DEGRADED|BCH_FORCE_IF_LOST; + + if (c->opts.degraded) + flags |= BCH_FORCE_IF_DEGRADED; + + if (!c->opts.degraded && + !c->opts.very_degraded) { + mutex_lock(&c->sb_lock); + + for (i = 0; i < c->disk_sb.sb->nr_devices; i++) { + if (!bch2_member_exists(c->disk_sb.sb, i)) + continue; + + ca = bch2_dev_locked(c, i); + + if (!bch2_dev_is_online(ca) && + (ca->mi.state == BCH_MEMBER_STATE_rw || + ca->mi.state == BCH_MEMBER_STATE_ro)) { + mutex_unlock(&c->sb_lock); + return false; + } + } + mutex_unlock(&c->sb_lock); + } + + return bch2_have_enough_devs(c, bch2_online_devs(c), flags, true); +} + int bch2_fs_start(struct bch_fs *c) { time64_t now = ktime_get_real_seconds(); @@ -1034,6 +1058,9 @@ int bch2_fs_start(struct bch_fs *c) print_mount_opts(c); + if (!bch2_fs_may_start(c)) + return -BCH_ERR_insufficient_devices_to_start; + down_write(&c->state_lock); mutex_lock(&c->sb_lock); @@ -1086,13 +1113,10 @@ int bch2_fs_start(struct bch_fs *c) wake_up(&c->ro_ref_wait); down_write(&c->state_lock); - if (c->opts.read_only) { + if (c->opts.read_only) bch2_fs_read_only(c); - } else { - ret = !test_bit(BCH_FS_rw, &c->flags) - ? bch2_fs_read_write(c) - : bch2_fs_read_write_late(c); - } + else if (!test_bit(BCH_FS_rw, &c->flags)) + ret = bch2_fs_read_write(c); up_write(&c->state_lock); err: @@ -1504,7 +1528,7 @@ static int bch2_dev_attach_bdev(struct bch_fs *c, struct bch_sb_handle *sb) printbuf_exit(&name); - rebalance_wakeup(c); + bch2_rebalance_wakeup(c); return 0; } @@ -1563,40 +1587,6 @@ bool bch2_dev_state_allowed(struct bch_fs *c, struct bch_dev *ca, } } -static bool bch2_fs_may_start(struct bch_fs *c) -{ - struct bch_dev *ca; - unsigned i, flags = 0; - - if (c->opts.very_degraded) - flags |= BCH_FORCE_IF_DEGRADED|BCH_FORCE_IF_LOST; - - if (c->opts.degraded) - flags |= BCH_FORCE_IF_DEGRADED; - - if (!c->opts.degraded && - !c->opts.very_degraded) { - mutex_lock(&c->sb_lock); - - for (i = 0; i < c->disk_sb.sb->nr_devices; i++) { - if (!bch2_member_exists(c->disk_sb.sb, i)) - continue; - - ca = bch2_dev_locked(c, i); - - if (!bch2_dev_is_online(ca) && - (ca->mi.state == BCH_MEMBER_STATE_rw || - ca->mi.state == BCH_MEMBER_STATE_ro)) { - mutex_unlock(&c->sb_lock); - return false; - } - } - mutex_unlock(&c->sb_lock); - } - - return bch2_have_enough_devs(c, bch2_online_devs(c), flags, true); -} - static void __bch2_dev_read_only(struct bch_fs *c, struct bch_dev *ca) { bch2_dev_io_ref_stop(ca, WRITE); @@ -1650,7 +1640,7 @@ int __bch2_dev_set_state(struct bch_fs *c, struct bch_dev *ca, if (new_state == BCH_MEMBER_STATE_rw) __bch2_dev_read_write(c, ca); - rebalance_wakeup(c); + bch2_rebalance_wakeup(c); return ret; } @@ -1767,7 +1757,8 @@ int bch2_dev_remove(struct bch_fs *c, struct bch_dev *ca, int flags) up_write(&c->state_lock); return 0; err: - if (ca->mi.state == BCH_MEMBER_STATE_rw && + if (test_bit(BCH_FS_rw, &c->flags) && + ca->mi.state == BCH_MEMBER_STATE_rw && !percpu_ref_is_zero(&ca->io_ref[READ])) __bch2_dev_read_write(c, ca); up_write(&c->state_lock); @@ -2231,11 +2222,6 @@ struct bch_fs *bch2_fs_open(char * const *devices, unsigned nr_devices, } up_write(&c->state_lock); - if (!bch2_fs_may_start(c)) { - ret = -BCH_ERR_insufficient_devices_to_start; - goto err_print; - } - if (!c->opts.nostart) { ret = bch2_fs_start(c); if (ret) diff --git a/fs/bcachefs/sysfs.c b/fs/bcachefs/sysfs.c index e5f003c29369..82ee333ddd21 100644 --- a/fs/bcachefs/sysfs.c +++ b/fs/bcachefs/sysfs.c @@ -654,11 +654,10 @@ static ssize_t sysfs_opt_store(struct bch_fs *c, bch2_set_rebalance_needs_scan(c, 0); if (v && id == Opt_rebalance_enabled) - rebalance_wakeup(c); + bch2_rebalance_wakeup(c); - if (v && id == Opt_copygc_enabled && - c->copygc_thread) - wake_up_process(c->copygc_thread); + if (v && id == Opt_copygc_enabled) + bch2_copygc_wakeup(c); if (id == Opt_discard && !ca) { mutex_lock(&c->sb_lock); diff --git a/fs/bcachefs/tests.c b/fs/bcachefs/tests.c index c265b102267a..782a05fe7656 100644 --- a/fs/bcachefs/tests.c +++ b/fs/bcachefs/tests.c @@ -342,6 +342,8 @@ static int test_iterate_slots_extents(struct bch_fs *c, u64 nr) */ static int test_peek_end(struct bch_fs *c, u64 nr) { + delete_test_keys(c); + struct btree_trans *trans = bch2_trans_get(c); struct btree_iter iter; struct bkey_s_c k; @@ -362,6 +364,8 @@ static int test_peek_end(struct bch_fs *c, u64 nr) static int test_peek_end_extents(struct bch_fs *c, u64 nr) { + delete_test_keys(c); + struct btree_trans *trans = bch2_trans_get(c); struct btree_iter iter; struct bkey_s_c k; diff --git a/fs/bcachefs/thread_with_file.c b/fs/bcachefs/thread_with_file.c index dea73bc1cb51..314a24d15d4e 100644 --- a/fs/bcachefs/thread_with_file.c +++ b/fs/bcachefs/thread_with_file.c @@ -455,8 +455,10 @@ ssize_t bch2_stdio_redirect_vprintf(struct stdio_redirect *stdio, bool nonblocki struct stdio_buf *buf = &stdio->output; unsigned long flags; ssize_t ret; - again: + if (stdio->done) + return -EPIPE; + spin_lock_irqsave(&buf->lock, flags); ret = bch2_darray_vprintf(&buf->buf, GFP_NOWAIT, fmt, args); spin_unlock_irqrestore(&buf->lock, flags); diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h index 6ba5071ab6dd..3e52c7f8ddd2 100644 --- a/fs/bcachefs/util.h +++ b/fs/bcachefs/util.h @@ -739,4 +739,42 @@ static inline void memcpy_swab(void *_dst, void *_src, size_t len) *--dst = *src++; } +#define set_flags(_map, _in, _out) \ +do { \ + unsigned _i; \ + \ + for (_i = 0; _i < ARRAY_SIZE(_map); _i++) \ + if ((_in) & (1 << _i)) \ + (_out) |= _map[_i]; \ + else \ + (_out) &= ~_map[_i]; \ +} while (0) + +#define map_flags(_map, _in) \ +({ \ + unsigned _out = 0; \ + \ + set_flags(_map, _in, _out); \ + _out; \ +}) + +#define map_flags_rev(_map, _in) \ +({ \ + unsigned _i, _out = 0; \ + \ + for (_i = 0; _i < ARRAY_SIZE(_map); _i++) \ + if ((_in) & _map[_i]) { \ + (_out) |= 1 << _i; \ + (_in) &= ~_map[_i]; \ + } \ + (_out); \ +}) + +#define map_defined(_map) \ +({ \ + unsigned _in = ~0; \ + \ + map_flags_rev(_map, _in); \ +}) + #endif /* _BCACHEFS_UTIL_H */ diff --git a/fs/bcachefs/xattr_format.h b/fs/bcachefs/xattr_format.h index c7916011ef34..67426e33d04e 100644 --- a/fs/bcachefs/xattr_format.h +++ b/fs/bcachefs/xattr_format.h @@ -13,7 +13,13 @@ struct bch_xattr { __u8 x_type; __u8 x_name_len; __le16 x_val_len; - __u8 x_name[] __counted_by(x_name_len); + /* + * x_name contains the name and value counted by + * x_name_len + x_val_len. The introduction of + * __counted_by(x_name_len) caused a false positive + * detection of an out of bounds write. + */ + __u8 x_name[]; } __packed __aligned(8); #endif /* _BCACHEFS_XATTR_FORMAT_H */ |