+
+int bch2_inum_opts_get(struct btree_trans *trans, subvol_inum inum, struct bch_io_opts *opts)
+{
+ struct bch_inode_unpacked inode;
+ int ret = lockrestart_do(trans, bch2_inode_find_by_inum_trans(trans, inum, &inode));
+
+ if (ret)
+ return ret;
+
+ bch2_inode_opts_get(opts, trans->c, &inode);
+ return 0;
+}
+
+int bch2_inode_rm_snapshot(struct btree_trans *trans, u64 inum, u32 snapshot)
+{
+ struct bch_fs *c = trans->c;
+ struct btree_iter iter = { NULL };
+ struct bkey_i_inode_generation delete;
+ struct bch_inode_unpacked inode_u;
+ struct bkey_s_c k;
+ int ret;
+
+ do {
+ ret = bch2_btree_delete_range_trans(trans, BTREE_ID_extents,
+ SPOS(inum, 0, snapshot),
+ SPOS(inum, U64_MAX, snapshot),
+ 0, NULL) ?:
+ bch2_btree_delete_range_trans(trans, BTREE_ID_dirents,
+ SPOS(inum, 0, snapshot),
+ SPOS(inum, U64_MAX, snapshot),
+ 0, NULL) ?:
+ bch2_btree_delete_range_trans(trans, BTREE_ID_xattrs,
+ SPOS(inum, 0, snapshot),
+ SPOS(inum, U64_MAX, snapshot),
+ 0, NULL);
+ } while (ret == -BCH_ERR_transaction_restart_nested);
+ if (ret)
+ goto err;
+retry:
+ bch2_trans_begin(trans);
+
+ k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
+ SPOS(0, inum, snapshot), BTREE_ITER_INTENT);
+ ret = bkey_err(k);
+ if (ret)
+ goto err;
+
+ if (!bkey_is_inode(k.k)) {
+ bch2_fs_inconsistent(c,
+ "inode %llu:%u not found when deleting",
+ inum, snapshot);
+ ret = -EIO;
+ goto err;
+ }
+
+ bch2_inode_unpack(k, &inode_u);
+
+ /* Subvolume root? */
+ if (inode_u.bi_subvol)
+ bch_warn(c, "deleting inode %llu marked as unlinked, but also a subvolume root!?", inode_u.bi_inum);
+
+ bkey_inode_generation_init(&delete.k_i);
+ delete.k.p = iter.pos;
+ delete.v.bi_generation = cpu_to_le32(inode_u.bi_generation + 1);
+
+ ret = bch2_trans_update(trans, &iter, &delete.k_i, 0) ?:
+ bch2_trans_commit(trans, NULL, NULL,
+ BCH_TRANS_COMMIT_no_enospc);
+err:
+ bch2_trans_iter_exit(trans, &iter);
+ if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
+ goto retry;
+
+ return ret ?: -BCH_ERR_transaction_restart_nested;
+}
+
+static int may_delete_deleted_inode(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bpos pos,
+ bool *need_another_pass)
+{
+ struct bch_fs *c = trans->c;
+ struct btree_iter inode_iter;
+ struct bkey_s_c k;
+ struct bch_inode_unpacked inode;
+ int ret;
+
+ k = bch2_bkey_get_iter(trans, &inode_iter, BTREE_ID_inodes, pos, BTREE_ITER_CACHED);
+ ret = bkey_err(k);
+ if (ret)
+ return ret;
+
+ ret = bkey_is_inode(k.k) ? 0 : -BCH_ERR_ENOENT_inode;
+ if (fsck_err_on(!bkey_is_inode(k.k), c,
+ deleted_inode_missing,
+ "nonexistent inode %llu:%u in deleted_inodes btree",
+ pos.offset, pos.snapshot))
+ goto delete;
+
+ ret = bch2_inode_unpack(k, &inode);
+ if (ret)
+ goto out;
+
+ if (fsck_err_on(S_ISDIR(inode.bi_mode), c,
+ deleted_inode_is_dir,
+ "directory %llu:%u in deleted_inodes btree",
+ pos.offset, pos.snapshot))
+ goto delete;
+
+ if (fsck_err_on(!(inode.bi_flags & BCH_INODE_unlinked), c,
+ deleted_inode_not_unlinked,
+ "non-deleted inode %llu:%u in deleted_inodes btree",
+ pos.offset, pos.snapshot))
+ goto delete;
+
+ if (c->sb.clean &&
+ !fsck_err(c,
+ deleted_inode_but_clean,
+ "filesystem marked as clean but have deleted inode %llu:%u",
+ pos.offset, pos.snapshot)) {
+ ret = 0;
+ goto out;
+ }
+
+ if (bch2_snapshot_is_internal_node(c, pos.snapshot)) {
+ struct bpos new_min_pos;
+
+ ret = bch2_propagate_key_to_snapshot_leaves(trans, inode_iter.btree_id, k, &new_min_pos);
+ if (ret)
+ goto out;
+
+ inode.bi_flags &= ~BCH_INODE_unlinked;
+
+ ret = bch2_inode_write_flags(trans, &inode_iter, &inode,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+ bch2_trans_commit(trans, NULL, NULL,
+ BCH_TRANS_COMMIT_no_enospc|
+ BCH_TRANS_COMMIT_lazy_rw);
+ bch_err_msg(c, ret, "clearing inode unlinked flag");
+ if (ret)
+ goto out;
+
+ /*
+ * We'll need another write buffer flush to pick up the new
+ * unlinked inodes in the snapshot leaves:
+ */
+ *need_another_pass = true;
+ goto out;
+ }
+
+ ret = 1;
+out:
+fsck_err:
+ bch2_trans_iter_exit(trans, &inode_iter);
+ return ret;
+delete:
+ ret = bch2_btree_bit_mod(trans, BTREE_ID_deleted_inodes, pos, false);
+ goto out;
+}
+
+int bch2_delete_dead_inodes(struct bch_fs *c)
+{
+ struct btree_trans *trans = bch2_trans_get(c);
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ bool need_another_pass;
+ int ret;
+again:
+ need_another_pass = false;
+
+ ret = bch2_btree_write_buffer_flush_sync(trans);
+ if (ret)
+ goto err;
+
+ /*
+ * Weird transaction restart handling here because on successful delete,
+ * bch2_inode_rm_snapshot() will return a nested transaction restart,
+ * but we can't retry because the btree write buffer won't have been
+ * flushed and we'd spin:
+ */
+ for_each_btree_key(trans, iter, BTREE_ID_deleted_inodes, POS_MIN,
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+ ret = lockrestart_do(trans, may_delete_deleted_inode(trans, &iter, k.k->p,
+ &need_another_pass));
+ if (ret < 0)
+ break;
+
+ if (ret) {
+ if (!test_bit(BCH_FS_RW, &c->flags)) {
+ bch2_trans_unlock(trans);
+ bch2_fs_lazy_rw(c);
+ }
+
+ bch_verbose(c, "deleting unlinked inode %llu:%u", k.k->p.offset, k.k->p.snapshot);
+
+ ret = bch2_inode_rm_snapshot(trans, k.k->p.offset, k.k->p.snapshot);
+ if (ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart))
+ break;
+ }
+ }
+ bch2_trans_iter_exit(trans, &iter);
+
+ if (!ret && need_another_pass)
+ goto again;
+err:
+ bch2_trans_put(trans);
+
+ return ret;
+}