+ int ret;
+
+ snapshots_seen_init(&s);
+
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
+ POS_MIN,
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+ NULL, NULL, BCH_TRANS_COMMIT_lazy_rw|BCH_TRANS_COMMIT_no_enospc,
+ check_inode(trans, &iter, k, &prev, &s, full));
+
+ snapshots_seen_exit(&s);
+ bch2_trans_put(trans);
+ bch_err_fn(c, ret);
+ return ret;
+}
+
+static struct bkey_s_c_dirent dirent_get_by_pos(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bpos pos)
+{
+ return bch2_bkey_get_iter_typed(trans, iter, BTREE_ID_dirents, pos, 0, dirent);
+}
+
+static bool inode_points_to_dirent(struct bch_inode_unpacked *inode,
+ struct bkey_s_c_dirent d)
+{
+ return inode->bi_dir == d.k->p.inode &&
+ inode->bi_dir_offset == d.k->p.offset;
+}
+
+static bool dirent_points_to_inode(struct bkey_s_c_dirent d,
+ struct bch_inode_unpacked *inode)
+{
+ return d.v->d_type == DT_SUBVOL
+ ? le32_to_cpu(d.v->d_child_subvol) == inode->bi_subvol
+ : le64_to_cpu(d.v->d_inum) == inode->bi_inum;
+}
+
+static int inode_backpointer_exists(struct btree_trans *trans,
+ struct bch_inode_unpacked *inode,
+ u32 snapshot)
+{
+ struct btree_iter iter;
+ struct bkey_s_c_dirent d;
+ int ret;
+
+ d = dirent_get_by_pos(trans, &iter,
+ SPOS(inode->bi_dir, inode->bi_dir_offset, snapshot));
+ ret = bkey_err(d);
+ if (ret)
+ return bch2_err_matches(ret, ENOENT) ? 0 : ret;
+
+ ret = dirent_points_to_inode(d, inode);
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
+}
+
+static int check_i_sectors(struct btree_trans *trans, struct inode_walker *w)
+{
+ struct bch_fs *c = trans->c;
+ struct inode_walker_entry *i;
+ u32 restart_count = trans->restart_count;
+ int ret = 0;
+ s64 count2;
+
+ darray_for_each(w->inodes, i) {
+ if (i->inode.bi_sectors == i->count)
+ continue;
+
+ count2 = bch2_count_inode_sectors(trans, w->last_pos.inode, i->snapshot);
+
+ if (w->recalculate_sums)
+ i->count = count2;
+
+ if (i->count != count2) {
+ bch_err(c, "fsck counted i_sectors wrong for inode %llu:%u: got %llu should be %llu",
+ w->last_pos.inode, i->snapshot, i->count, count2);
+ return -BCH_ERR_internal_fsck_err;
+ }
+
+ if (fsck_err_on(!(i->inode.bi_flags & BCH_INODE_i_sectors_dirty),
+ c, inode_i_sectors_wrong,
+ "inode %llu:%u has incorrect i_sectors: got %llu, should be %llu",
+ w->last_pos.inode, i->snapshot,
+ i->inode.bi_sectors, i->count)) {
+ i->inode.bi_sectors = i->count;
+ ret = fsck_write_inode(trans, &i->inode, i->snapshot);
+ if (ret)
+ break;
+ }
+ }
+fsck_err:
+ bch_err_fn(c, ret);
+ return ret ?: trans_was_restarted(trans, restart_count);
+}
+
+struct extent_end {
+ u32 snapshot;
+ u64 offset;
+ struct snapshots_seen seen;
+};
+
+struct extent_ends {
+ struct bpos last_pos;
+ DARRAY(struct extent_end) e;
+};
+
+static void extent_ends_reset(struct extent_ends *extent_ends)
+{
+ struct extent_end *i;
+
+ darray_for_each(extent_ends->e, i)
+ snapshots_seen_exit(&i->seen);
+
+ extent_ends->e.nr = 0;
+}
+
+static void extent_ends_exit(struct extent_ends *extent_ends)
+{
+ extent_ends_reset(extent_ends);
+ darray_exit(&extent_ends->e);
+}
+
+static void extent_ends_init(struct extent_ends *extent_ends)
+{
+ memset(extent_ends, 0, sizeof(*extent_ends));
+}
+
+static int extent_ends_at(struct bch_fs *c,
+ struct extent_ends *extent_ends,
+ struct snapshots_seen *seen,
+ struct bkey_s_c k)
+{
+ struct extent_end *i, n = (struct extent_end) {
+ .offset = k.k->p.offset,
+ .snapshot = k.k->p.snapshot,
+ .seen = *seen,
+ };
+
+ n.seen.ids.data = kmemdup(seen->ids.data,
+ sizeof(seen->ids.data[0]) * seen->ids.size,
+ GFP_KERNEL);
+ if (!n.seen.ids.data)
+ return -BCH_ERR_ENOMEM_fsck_extent_ends_at;
+
+ darray_for_each(extent_ends->e, i) {
+ if (i->snapshot == k.k->p.snapshot) {
+ snapshots_seen_exit(&i->seen);
+ *i = n;
+ return 0;
+ }
+
+ if (i->snapshot >= k.k->p.snapshot)
+ break;
+ }
+
+ return darray_insert_item(&extent_ends->e, i - extent_ends->e.data, n);
+}
+
+static int overlapping_extents_found(struct btree_trans *trans,
+ enum btree_id btree,
+ struct bpos pos1, struct snapshots_seen *pos1_seen,
+ struct bkey pos2,
+ bool *fixed,
+ struct extent_end *extent_end)
+{
+ struct bch_fs *c = trans->c;
+ struct printbuf buf = PRINTBUF;
+ struct btree_iter iter1, iter2 = { NULL };
+ struct bkey_s_c k1, k2;
+ int ret;
+
+ BUG_ON(bkey_le(pos1, bkey_start_pos(&pos2)));
+
+ bch2_trans_iter_init(trans, &iter1, btree, pos1,
+ BTREE_ITER_ALL_SNAPSHOTS|
+ BTREE_ITER_NOT_EXTENTS);
+ k1 = bch2_btree_iter_peek_upto(&iter1, POS(pos1.inode, U64_MAX));
+ ret = bkey_err(k1);
+ if (ret)
+ goto err;
+
+ prt_str(&buf, "\n ");
+ bch2_bkey_val_to_text(&buf, c, k1);
+
+ if (!bpos_eq(pos1, k1.k->p)) {
+ prt_str(&buf, "\n wanted\n ");
+ bch2_bpos_to_text(&buf, pos1);
+ prt_str(&buf, "\n ");
+ bch2_bkey_to_text(&buf, &pos2);
+
+ bch_err(c, "%s: error finding first overlapping extent when repairing, got%s",
+ __func__, buf.buf);
+ ret = -BCH_ERR_internal_fsck_err;
+ goto err;
+ }
+
+ bch2_trans_copy_iter(&iter2, &iter1);
+
+ while (1) {
+ bch2_btree_iter_advance(&iter2);
+
+ k2 = bch2_btree_iter_peek_upto(&iter2, POS(pos1.inode, U64_MAX));
+ ret = bkey_err(k2);
+ if (ret)
+ goto err;
+
+ if (bpos_ge(k2.k->p, pos2.p))
+ break;
+ }
+
+ prt_str(&buf, "\n ");
+ bch2_bkey_val_to_text(&buf, c, k2);
+
+ if (bpos_gt(k2.k->p, pos2.p) ||
+ pos2.size != k2.k->size) {
+ bch_err(c, "%s: error finding seconding overlapping extent when repairing%s",
+ __func__, buf.buf);
+ ret = -BCH_ERR_internal_fsck_err;
+ goto err;
+ }
+
+ prt_printf(&buf, "\n overwriting %s extent",
+ pos1.snapshot >= pos2.p.snapshot ? "first" : "second");
+
+ if (fsck_err(c, extent_overlapping,
+ "overlapping extents%s", buf.buf)) {
+ struct btree_iter *old_iter = &iter1;
+ struct disk_reservation res = { 0 };
+
+ if (pos1.snapshot < pos2.p.snapshot) {
+ old_iter = &iter2;
+ swap(k1, k2);
+ }
+
+ trans->extra_journal_res += bch2_bkey_sectors_compressed(k2);
+
+ ret = bch2_trans_update_extent_overwrite(trans, old_iter,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE,
+ k1, k2) ?:
+ bch2_trans_commit(trans, &res, NULL,
+ BCH_TRANS_COMMIT_lazy_rw|BCH_TRANS_COMMIT_no_enospc);
+ bch2_disk_reservation_put(c, &res);
+
+ if (ret)
+ goto err;
+
+ *fixed = true;
+
+ if (pos1.snapshot == pos2.p.snapshot) {
+ /*
+ * We overwrote the first extent, and did the overwrite
+ * in the same snapshot:
+ */
+ extent_end->offset = bkey_start_offset(&pos2);
+ } else if (pos1.snapshot > pos2.p.snapshot) {
+ /*
+ * We overwrote the first extent in pos2's snapshot:
+ */
+ ret = snapshots_seen_add_inorder(c, pos1_seen, pos2.p.snapshot);
+ } else {
+ /*
+ * We overwrote the second extent - restart
+ * check_extent() from the top:
+ */
+ ret = -BCH_ERR_transaction_restart_nested;
+ }
+ }
+fsck_err:
+err:
+ bch2_trans_iter_exit(trans, &iter2);
+ bch2_trans_iter_exit(trans, &iter1);
+ printbuf_exit(&buf);
+ return ret;
+}
+
+static int check_overlapping_extents(struct btree_trans *trans,
+ struct snapshots_seen *seen,
+ struct extent_ends *extent_ends,
+ struct bkey_s_c k,
+ u32 equiv,
+ struct btree_iter *iter,
+ bool *fixed)
+{
+ struct bch_fs *c = trans->c;
+ struct extent_end *i;
+ int ret = 0;
+
+ /* transaction restart, running again */
+ if (bpos_eq(extent_ends->last_pos, k.k->p))
+ return 0;
+
+ if (extent_ends->last_pos.inode != k.k->p.inode)
+ extent_ends_reset(extent_ends);
+
+ darray_for_each(extent_ends->e, i) {
+ if (i->offset <= bkey_start_offset(k.k))
+ continue;
+
+ if (!ref_visible2(c,
+ k.k->p.snapshot, seen,
+ i->snapshot, &i->seen))
+ continue;
+
+ ret = overlapping_extents_found(trans, iter->btree_id,
+ SPOS(iter->pos.inode,
+ i->offset,
+ i->snapshot),
+ &i->seen,
+ *k.k, fixed, i);
+ if (ret)
+ goto err;
+ }
+
+ ret = extent_ends_at(c, extent_ends, seen, k);
+ if (ret)
+ goto err;
+
+ extent_ends->last_pos = k.k->p;
+err:
+ return ret;
+}
+
+static int check_extent_overbig(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k)
+{
+ struct bch_fs *c = trans->c;
+ struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+ struct bch_extent_crc_unpacked crc;
+ const union bch_extent_entry *i;
+ unsigned encoded_extent_max_sectors = c->opts.encoded_extent_max >> 9;
+
+ bkey_for_each_crc(k.k, ptrs, crc, i)
+ if (crc_is_encoded(crc) &&
+ crc.uncompressed_size > encoded_extent_max_sectors) {
+ struct printbuf buf = PRINTBUF;
+
+ bch2_bkey_val_to_text(&buf, c, k);
+ bch_err(c, "overbig encoded extent, please report this:\n %s", buf.buf);
+ printbuf_exit(&buf);
+ }
+
+ return 0;
+}
+
+static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
+ struct inode_walker *inode,
+ struct snapshots_seen *s,
+ struct extent_ends *extent_ends)
+{
+ struct bch_fs *c = trans->c;
+ struct inode_walker_entry *i;
+ struct printbuf buf = PRINTBUF;
+ struct bpos equiv = k.k->p;
+ int ret = 0;
+
+ equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
+
+ ret = check_key_has_snapshot(trans, iter, k);
+ if (ret) {
+ ret = ret < 0 ? ret : 0;
+ goto out;
+ }
+
+ if (inode->last_pos.inode != k.k->p.inode) {
+ ret = check_i_sectors(trans, inode);
+ if (ret)
+ goto err;
+ }
+
+ i = walk_inode(trans, inode, equiv, k.k->type == KEY_TYPE_whiteout);
+ ret = PTR_ERR_OR_ZERO(i);
+ if (ret)
+ goto err;
+
+ ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
+ if (ret)
+ goto err;
+
+ if (k.k->type != KEY_TYPE_whiteout) {
+ if (fsck_err_on(!i, c, extent_in_missing_inode,
+ "extent in missing inode:\n %s",
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+ goto delete;
+
+ if (fsck_err_on(i &&
+ !S_ISREG(i->inode.bi_mode) &&
+ !S_ISLNK(i->inode.bi_mode),
+ c, extent_in_non_reg_inode,
+ "extent in non regular inode mode %o:\n %s",
+ i->inode.bi_mode,
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+ goto delete;
+
+ ret = check_overlapping_extents(trans, s, extent_ends, k,
+ equiv.snapshot, iter,
+ &inode->recalculate_sums);
+ if (ret)
+ goto err;
+ }
+
+ /*
+ * Check inodes in reverse order, from oldest snapshots to newest,
+ * starting from the inode that matches this extent's snapshot. If we
+ * didn't have one, iterate over all inodes:
+ */
+ if (!i)
+ i = inode->inodes.data + inode->inodes.nr - 1;
+
+ for (;
+ inode->inodes.data && i >= inode->inodes.data;
+ --i) {
+ if (i->snapshot > equiv.snapshot ||
+ !key_visible_in_snapshot(c, s, i->snapshot, equiv.snapshot))
+ continue;
+
+ if (k.k->type != KEY_TYPE_whiteout) {
+ if (fsck_err_on(!(i->inode.bi_flags & BCH_INODE_i_size_dirty) &&
+ k.k->p.offset > round_up(i->inode.bi_size, block_bytes(c)) >> 9 &&
+ !bkey_extent_is_reservation(k),
+ c, extent_past_end_of_inode,
+ "extent type past end of inode %llu:%u, i_size %llu\n %s",
+ i->inode.bi_inum, i->snapshot, i->inode.bi_size,
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+ struct btree_iter iter2;
+
+ bch2_trans_copy_iter(&iter2, iter);
+ bch2_btree_iter_set_snapshot(&iter2, i->snapshot);
+ ret = bch2_btree_iter_traverse(&iter2) ?:
+ bch2_btree_delete_at(trans, &iter2,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+ bch2_trans_iter_exit(trans, &iter2);
+ if (ret)
+ goto err;
+
+ iter->k.type = KEY_TYPE_whiteout;
+ }
+
+ if (bkey_extent_is_allocation(k.k))
+ i->count += k.k->size;
+ }
+
+ i->seen_this_pos = true;
+ }
+out:
+err:
+fsck_err:
+ printbuf_exit(&buf);
+ bch_err_fn(c, ret);
+ return ret;
+delete:
+ ret = bch2_btree_delete_at(trans, iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+ goto out;
+}
+
+/*
+ * Walk extents: verify that extents have a corresponding S_ISREG inode, and
+ * that i_size an i_sectors are consistent
+ */
+int bch2_check_extents(struct bch_fs *c)
+{
+ struct inode_walker w = inode_walker_init();
+ struct snapshots_seen s;
+ struct btree_trans *trans = bch2_trans_get(c);
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ struct extent_ends extent_ends;
+ struct disk_reservation res = { 0 };
+ int ret = 0;
+
+ snapshots_seen_init(&s);
+ extent_ends_init(&extent_ends);
+
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_extents,
+ POS(BCACHEFS_ROOT_INO, 0),
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+ &res, NULL,
+ BCH_TRANS_COMMIT_lazy_rw|BCH_TRANS_COMMIT_no_enospc, ({
+ bch2_disk_reservation_put(c, &res);
+ check_extent(trans, &iter, k, &w, &s, &extent_ends) ?:
+ check_extent_overbig(trans, &iter, k);
+ })) ?:
+ check_i_sectors(trans, &w);
+
+ bch2_disk_reservation_put(c, &res);
+ extent_ends_exit(&extent_ends);
+ inode_walker_exit(&w);
+ snapshots_seen_exit(&s);
+ bch2_trans_put(trans);
+
+ bch_err_fn(c, ret);
+ return ret;
+}
+
+int bch2_check_indirect_extents(struct bch_fs *c)
+{
+ struct btree_trans *trans = bch2_trans_get(c);
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ struct disk_reservation res = { 0 };
+ int ret = 0;
+
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_reflink,
+ POS_MIN,
+ BTREE_ITER_PREFETCH, k,
+ &res, NULL,
+ BCH_TRANS_COMMIT_lazy_rw|BCH_TRANS_COMMIT_no_enospc, ({
+ bch2_disk_reservation_put(c, &res);
+ check_extent_overbig(trans, &iter, k);
+ }));
+
+ bch2_disk_reservation_put(c, &res);
+ bch2_trans_put(trans);
+
+ bch_err_fn(c, ret);
+ return ret;
+}
+
+static int check_subdir_count(struct btree_trans *trans, struct inode_walker *w)
+{
+ struct bch_fs *c = trans->c;
+ struct inode_walker_entry *i;
+ u32 restart_count = trans->restart_count;
+ int ret = 0;
+ s64 count2;
+
+ darray_for_each(w->inodes, i) {
+ if (i->inode.bi_nlink == i->count)
+ continue;
+
+ count2 = bch2_count_subdirs(trans, w->last_pos.inode, i->snapshot);
+ if (count2 < 0)
+ return count2;
+
+ if (i->count != count2) {
+ bch_err(c, "fsck counted subdirectories wrong: got %llu should be %llu",
+ i->count, count2);
+ i->count = count2;
+ if (i->inode.bi_nlink == i->count)
+ continue;
+ }
+
+ if (fsck_err_on(i->inode.bi_nlink != i->count,
+ c, inode_dir_wrong_nlink,
+ "directory %llu:%u with wrong i_nlink: got %u, should be %llu",
+ w->last_pos.inode, i->snapshot, i->inode.bi_nlink, i->count)) {
+ i->inode.bi_nlink = i->count;
+ ret = fsck_write_inode(trans, &i->inode, i->snapshot);
+ if (ret)
+ break;
+ }
+ }
+fsck_err:
+ bch_err_fn(c, ret);
+ return ret ?: trans_was_restarted(trans, restart_count);
+}
+
+static int check_dirent_target(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bkey_s_c_dirent d,
+ struct bch_inode_unpacked *target,
+ u32 target_snapshot)
+{
+ struct bch_fs *c = trans->c;
+ struct bkey_i_dirent *n;
+ bool backpointer_exists = true;
+ struct printbuf buf = PRINTBUF;
+ int ret = 0;
+
+ if (!target->bi_dir &&
+ !target->bi_dir_offset) {
+ target->bi_dir = d.k->p.inode;
+ target->bi_dir_offset = d.k->p.offset;
+
+ ret = __write_inode(trans, target, target_snapshot);
+ if (ret)
+ goto err;
+ }
+
+ if (!inode_points_to_dirent(target, d)) {
+ ret = inode_backpointer_exists(trans, target, d.k->p.snapshot);
+ if (ret < 0)
+ goto err;
+
+ backpointer_exists = ret;
+ ret = 0;
+
+ if (fsck_err_on(S_ISDIR(target->bi_mode) && backpointer_exists,
+ c, inode_dir_multiple_links,
+ "directory %llu with multiple links",
+ target->bi_inum)) {
+ ret = __remove_dirent(trans, d.k->p);
+ goto out;
+ }
+
+ if (fsck_err_on(backpointer_exists && !target->bi_nlink,
+ c, inode_multiple_links_but_nlink_0,
+ "inode %llu type %s has multiple links but i_nlink 0",
+ target->bi_inum, bch2_d_types[d.v->d_type])) {
+ target->bi_nlink++;
+ target->bi_flags &= ~BCH_INODE_unlinked;
+
+ ret = __write_inode(trans, target, target_snapshot);
+ if (ret)
+ goto err;
+ }
+
+ if (fsck_err_on(!backpointer_exists,
+ c, inode_wrong_backpointer,
+ "inode %llu:%u has wrong backpointer:\n"
+ "got %llu:%llu\n"
+ "should be %llu:%llu",
+ target->bi_inum, target_snapshot,
+ target->bi_dir,
+ target->bi_dir_offset,
+ d.k->p.inode,
+ d.k->p.offset)) {
+ target->bi_dir = d.k->p.inode;
+ target->bi_dir_offset = d.k->p.offset;
+
+ ret = __write_inode(trans, target, target_snapshot);
+ if (ret)
+ goto err;
+ }
+ }
+
+ if (fsck_err_on(d.v->d_type != inode_d_type(target),
+ c, dirent_d_type_wrong,
+ "incorrect d_type: got %s, should be %s:\n%s",
+ bch2_d_type_str(d.v->d_type),
+ bch2_d_type_str(inode_d_type(target)),
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, d.s_c), buf.buf))) {
+ n = bch2_trans_kmalloc(trans, bkey_bytes(d.k));
+ ret = PTR_ERR_OR_ZERO(n);
+ if (ret)
+ goto err;
+
+ bkey_reassemble(&n->k_i, d.s_c);
+ n->v.d_type = inode_d_type(target);
+
+ ret = bch2_trans_update(trans, iter, &n->k_i, 0);
+ if (ret)
+ goto err;
+
+ d = dirent_i_to_s_c(n);
+ }
+
+ if (d.v->d_type == DT_SUBVOL &&
+ target->bi_parent_subvol != le32_to_cpu(d.v->d_parent_subvol) &&
+ (c->sb.version < bcachefs_metadata_version_subvol_dirent ||
+ fsck_err(c, dirent_d_parent_subvol_wrong,
+ "dirent has wrong d_parent_subvol field: got %u, should be %u",
+ le32_to_cpu(d.v->d_parent_subvol),
+ target->bi_parent_subvol))) {
+ n = bch2_trans_kmalloc(trans, bkey_bytes(d.k));
+ ret = PTR_ERR_OR_ZERO(n);
+ if (ret)
+ goto err;
+
+ bkey_reassemble(&n->k_i, d.s_c);
+ n->v.d_parent_subvol = cpu_to_le32(target->bi_parent_subvol);
+
+ ret = bch2_trans_update(trans, iter, &n->k_i, 0);
+ if (ret)
+ goto err;
+
+ d = dirent_i_to_s_c(n);
+ }
+out:
+err:
+fsck_err:
+ printbuf_exit(&buf);
+ bch_err_fn(c, ret);
+ return ret;
+}
+
+static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
+ struct bch_hash_info *hash_info,
+ struct inode_walker *dir,
+ struct inode_walker *target,
+ struct snapshots_seen *s)
+{
+ struct bch_fs *c = trans->c;
+ struct bkey_s_c_dirent d;
+ struct inode_walker_entry *i;
+ struct printbuf buf = PRINTBUF;
+ struct bpos equiv;
+ int ret = 0;
+
+ ret = check_key_has_snapshot(trans, iter, k);
+ if (ret) {
+ ret = ret < 0 ? ret : 0;
+ goto out;
+ }
+
+ equiv = k.k->p;
+ equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
+
+ ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
+ if (ret)
+ goto err;
+
+ if (k.k->type == KEY_TYPE_whiteout)
+ goto out;
+
+ if (dir->last_pos.inode != k.k->p.inode) {
+ ret = check_subdir_count(trans, dir);
+ if (ret)
+ goto err;
+ }
+
+ BUG_ON(!iter->path->should_be_locked);
+
+ i = walk_inode(trans, dir, equiv, k.k->type == KEY_TYPE_whiteout);
+ ret = PTR_ERR_OR_ZERO(i);
+ if (ret < 0)
+ goto err;
+
+ if (dir->first_this_inode && dir->inodes.nr)
+ *hash_info = bch2_hash_info_init(c, &dir->inodes.data[0].inode);
+ dir->first_this_inode = false;
+
+ if (fsck_err_on(!i, c, dirent_in_missing_dir_inode,
+ "dirent in nonexisting directory:\n%s",
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+ ret = bch2_btree_delete_at(trans, iter,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+ goto out;
+ }
+
+ if (!i)
+ goto out;
+
+ if (fsck_err_on(!S_ISDIR(i->inode.bi_mode),
+ c, dirent_in_non_dir_inode,
+ "dirent in non directory inode type %s:\n%s",
+ bch2_d_type_str(inode_d_type(&i->inode)),
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+ ret = bch2_btree_delete_at(trans, iter, 0);
+ goto out;
+ }
+
+ ret = hash_check_key(trans, bch2_dirent_hash_desc, hash_info, iter, k);
+ if (ret < 0)
+ goto err;
+ if (ret) {
+ /* dirent has been deleted */
+ ret = 0;
+ goto out;
+ }
+
+ if (k.k->type != KEY_TYPE_dirent)
+ goto out;
+
+ d = bkey_s_c_to_dirent(k);
+
+ if (d.v->d_type == DT_SUBVOL) {
+ struct bch_inode_unpacked subvol_root;
+ u32 target_subvol = le32_to_cpu(d.v->d_child_subvol);
+ u32 target_snapshot;
+ u64 target_inum;
+
+ ret = __subvol_lookup(trans, target_subvol,
+ &target_snapshot, &target_inum);
+ if (ret && !bch2_err_matches(ret, ENOENT))
+ goto err;
+
+ if (fsck_err_on(ret, c, dirent_to_missing_subvol,
+ "dirent points to missing subvolume %u",
+ le32_to_cpu(d.v->d_child_subvol))) {
+ ret = __remove_dirent(trans, d.k->p);
+ goto err;
+ }
+
+ ret = __lookup_inode(trans, target_inum,
+ &subvol_root, &target_snapshot);
+ if (ret && !bch2_err_matches(ret, ENOENT))
+ goto err;
+
+ if (fsck_err_on(ret, c, subvol_to_missing_root,
+ "subvolume %u points to missing subvolume root %llu",
+ target_subvol,
+ target_inum)) {
+ bch_err(c, "repair not implemented yet");
+ ret = -EINVAL;
+ goto err;
+ }
+
+ if (fsck_err_on(subvol_root.bi_subvol != target_subvol,
+ c, subvol_root_wrong_bi_subvol,
+ "subvol root %llu has wrong bi_subvol field: got %u, should be %u",
+ target_inum,
+ subvol_root.bi_subvol, target_subvol)) {
+ subvol_root.bi_subvol = target_subvol;
+ ret = __write_inode(trans, &subvol_root, target_snapshot);
+ if (ret)
+ goto err;
+ }
+
+ ret = check_dirent_target(trans, iter, d, &subvol_root,
+ target_snapshot);
+ if (ret)
+ goto err;
+ } else {
+ ret = __get_visible_inodes(trans, target, s, le64_to_cpu(d.v->d_inum));
+ if (ret)
+ goto err;
+
+ if (fsck_err_on(!target->inodes.nr,
+ c, dirent_to_missing_inode,
+ "dirent points to missing inode: (equiv %u)\n%s",
+ equiv.snapshot,
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k),
+ buf.buf))) {
+ ret = __remove_dirent(trans, d.k->p);
+ if (ret)
+ goto err;
+ }
+
+ darray_for_each(target->inodes, i) {
+ ret = check_dirent_target(trans, iter, d,
+ &i->inode, i->snapshot);
+ if (ret)
+ goto err;
+ }
+ }
+
+ if (d.v->d_type == DT_DIR)
+ for_each_visible_inode(c, s, dir, equiv.snapshot, i)
+ i->count++;
+
+out:
+err:
+fsck_err:
+ printbuf_exit(&buf);
+ bch_err_fn(c, ret);
+ return ret;
+}
+
+/*
+ * Walk dirents: verify that they all have a corresponding S_ISDIR inode,
+ * validate d_type
+ */
+int bch2_check_dirents(struct bch_fs *c)
+{
+ struct inode_walker dir = inode_walker_init();
+ struct inode_walker target = inode_walker_init();
+ struct snapshots_seen s;
+ struct bch_hash_info hash_info;
+ struct btree_trans *trans = bch2_trans_get(c);
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ int ret = 0;
+
+ snapshots_seen_init(&s);
+
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_dirents,
+ POS(BCACHEFS_ROOT_INO, 0),
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+ k,
+ NULL, NULL,
+ BCH_TRANS_COMMIT_lazy_rw|BCH_TRANS_COMMIT_no_enospc,
+ check_dirent(trans, &iter, k, &hash_info, &dir, &target, &s));
+
+ bch2_trans_put(trans);
+ snapshots_seen_exit(&s);
+ inode_walker_exit(&dir);
+ inode_walker_exit(&target);
+ bch_err_fn(c, ret);
+ return ret;
+}
+
+static int check_xattr(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
+ struct bch_hash_info *hash_info,
+ struct inode_walker *inode)
+{
+ struct bch_fs *c = trans->c;
+ struct inode_walker_entry *i;
+ int ret;
+
+ ret = check_key_has_snapshot(trans, iter, k);
+ if (ret)
+ return ret;
+
+ i = walk_inode(trans, inode, k.k->p, k.k->type == KEY_TYPE_whiteout);
+ ret = PTR_ERR_OR_ZERO(i);
+ if (ret)
+ return ret;