netfilter: nf_tables: revisit chain/object refcounting from elements

Andreas reports that the following incremental update using our commit
protocol doesn't work.

# nft -f incremental-update.nft
delete element ip filter client_to_any { 10.180.86.22 : goto CIn_1 }
delete chain ip filter CIn_1
... Error: Could not process rule: Device or resource busy

The existing code is not well-integrated into the commit phase protocol,
since element deletions do not result in refcount decrement from the
preparation phase. This results in bogus EBUSY errors like the one
above.

Two new functions come with this patch:

* nft_set_elem_activate() function is used from the abort path, to
restore the set element refcounting on objects that occurred from
the preparation phase.

* nft_set_elem_deactivate() that is called from nft_del_setelem() to
decrement set element refcounting on objects from the preparation
phase in the commit protocol.

The nft_data_uninit() has been renamed to nft_data_release() since this
function does not uninitialize any data store in the data register,
instead just releases the references to objects. Moreover, a new
function nft_data_hold() has been introduced to be used from
nft_set_elem_activate().

Reported-by: Andreas Schultz <[e-mail]>
Signed-off-by: Pablo Neira Ayuso <[e-mail]>

include/net/netfilter/nf_tables.h

···
 int nft_data_init(const struct nft_ctx *ctx,
 		  struct nft_data *data, unsigned int size,
 		  struct nft_data_desc *desc, const struct nlattr *nla);
-void nft_data_uninit(const struct nft_data *data, enum nft_data_types type);
+void nft_data_release(const struct nft_data *data, enum nft_data_types type);
 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data,
 		  enum nft_data_types type, unsigned int len);
 

net/netfilter/nf_tables_api.c

···
 {
 	struct nft_set_ext *ext = nft_set_elem_ext(set, elem);
 
-	nft_data_uninit(nft_set_ext_key(ext), NFT_DATA_VALUE);
+	nft_data_release(nft_set_ext_key(ext), NFT_DATA_VALUE);
 	if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
-		nft_data_uninit(nft_set_ext_data(ext), set->dtype);
+		nft_data_release(nft_set_ext_data(ext), set->dtype);
 	if (destroy_expr && nft_set_ext_exists(ext, NFT_SET_EXT_EXPR))
 		nf_tables_expr_destroy(NULL, nft_set_ext_expr(ext));
 	if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF))
···
 	kfree(elem);
 }
 EXPORT_SYMBOL_GPL(nft_set_elem_destroy);
+
+/* Only called from commit path, nft_set_elem_deactivate() already deals with
+ * the refcounting from the preparation phase.
+ */
+static void nf_tables_set_elem_destroy(const struct nft_set *set, void *elem)
+{
+	struct nft_set_ext *ext = nft_set_elem_ext(set, elem);
+
+	if (nft_set_ext_exists(ext, NFT_SET_EXT_EXPR))
+		nf_tables_expr_destroy(NULL, nft_set_ext_expr(ext));
+	kfree(elem);
+}
 
 static int nft_setelem_parse_flags(const struct nft_set *set,
 				   const struct nlattr *attr, u32 *flags)
···
 	kfree(elem.priv);
 err3:
 	if (nla[NFTA_SET_ELEM_DATA] != NULL)
-		nft_data_uninit(&data, d2.type);
+		nft_data_release(&data, d2.type);
 err2:
-	nft_data_uninit(&elem.key.val, d1.type);
+	nft_data_release(&elem.key.val, d1.type);
 err1:
 	return err;
 }
···
 			break;
 	}
 	return err;
+}
+
+/**
+ *	nft_data_hold - hold a nft_data item
+ *
+ *	@data: struct nft_data to release
+ *	@type: type of data
+ *
+ *	Hold a nft_data item. NFT_DATA_VALUE types can be silently discarded,
+ *	NFT_DATA_VERDICT bumps the reference to chains in case of NFT_JUMP and
+ *	NFT_GOTO verdicts. This function must be called on active data objects
+ *	from the second phase of the commit protocol.
+ */
+static void nft_data_hold(const struct nft_data *data, enum nft_data_types type)
+{
+	if (type == NFT_DATA_VERDICT) {
+		switch (data->verdict.code) {
+		case NFT_JUMP:
+		case NFT_GOTO:
+			data->verdict.chain->use++;
+			break;
+		}
+	}
+}
+
+static void nft_set_elem_activate(const struct net *net,
+				  const struct nft_set *set,
+				  struct nft_set_elem *elem)
+{
+	const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
+
+	if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
+		nft_data_hold(nft_set_ext_data(ext), set->dtype);
+	if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF))
+		(*nft_set_ext_obj(ext))->use++;
+}
+
+static void nft_set_elem_deactivate(const struct net *net,
+				    const struct nft_set *set,
+				    struct nft_set_elem *elem)
+{
+	const struct nft_set_ext *ext = nft_set_elem_ext(set, elem->priv);
+
+	if (nft_set_ext_exists(ext, NFT_SET_EXT_DATA))
+		nft_data_release(nft_set_ext_data(ext), set->dtype);
+	if (nft_set_ext_exists(ext, NFT_SET_EXT_OBJREF))
+		(*nft_set_ext_obj(ext))->use--;
 }
 
 static int nft_del_setelem(struct nft_ctx *ctx, struct nft_set *set,
···
 	kfree(elem.priv);
 	elem.priv = priv;
 
+	nft_set_elem_deactivate(ctx->net, set, &elem);
+
 	nft_trans_elem(trans) = elem;
 	list_add_tail(&trans->list, &ctx->net->nft.commit_list);
 	return 0;
···
 err3:
 	kfree(elem.priv);
 err2:
-	nft_data_uninit(&elem.key.val, desc.type);
+	nft_data_release(&elem.key.val, desc.type);
 err1:
 	return err;
 }
···
 		nft_set_destroy(nft_trans_set(trans));
 		break;
 	case NFT_MSG_DELSETELEM:
-		nft_set_elem_destroy(nft_trans_elem_set(trans),
-				     nft_trans_elem(trans).priv, true);
+		nf_tables_set_elem_destroy(nft_trans_elem_set(trans),
+					   nft_trans_elem(trans).priv);
 		break;
 	case NFT_MSG_DELOBJ:
 		nft_obj_destroy(nft_trans_obj(trans));
···
 		case NFT_MSG_DELSETELEM:
 			te = (struct nft_trans_elem *)trans->data;
 
+			nft_set_elem_activate(net, te->set, &te->elem);
 			te->set->ops->activate(net, te->set, &te->elem);
 			te->set->ndeact--;
 
···
 EXPORT_SYMBOL_GPL(nft_data_init);
 
 /**
- *	nft_data_uninit - release a nft_data item
+ *	nft_data_release - release a nft_data item
  *
  *	@data: struct nft_data to release
  *	@type: type of data
···
  *	Release a nft_data item. NFT_DATA_VALUE types can be silently discarded,
  *	all others need to be released by calling this function.
  */
-void nft_data_uninit(const struct nft_data *data, enum nft_data_types type)
+void nft_data_release(const struct nft_data *data, enum nft_data_types type)
 {
 	if (type < NFT_DATA_VERDICT)
 		return;
···
 		WARN_ON(1);
 	}
 }
-EXPORT_SYMBOL_GPL(nft_data_uninit);
+EXPORT_SYMBOL_GPL(nft_data_release);
 
 int nft_data_dump(struct sk_buff *skb, int attr, const struct nft_data *data,
 		  enum nft_data_types type, unsigned int len)

net/netfilter/nft_bitwise.c

···
 
 	return 0;
 err2:
-	nft_data_uninit(&priv->xor, d2.type);
+	nft_data_release(&priv->xor, d2.type);
 err1:
-	nft_data_uninit(&priv->mask, d1.type);
+	nft_data_release(&priv->mask, d1.type);
 	return err;
 }
 

net/netfilter/nft_cmp.c

···
 
 	return &nft_cmp_ops;
 err1:
-	nft_data_uninit(&data, desc.type);
+	nft_data_release(&data, desc.type);
 	return ERR_PTR(-EINVAL);
 }
 

net/netfilter/nft_immediate.c

···
 	return 0;
 
 err1:
-	nft_data_uninit(&priv->data, desc.type);
+	nft_data_release(&priv->data, desc.type);
 	return err;
 }
 
···
 				  const struct nft_expr *expr)
 {
 	const struct nft_immediate_expr *priv = nft_expr_priv(expr);
-	return nft_data_uninit(&priv->data, nft_dreg_to_type(priv->dreg));
+
+	return nft_data_release(&priv->data, nft_dreg_to_type(priv->dreg));
 }
 
 static int nft_immediate_dump(struct sk_buff *skb, const struct nft_expr *expr)

net/netfilter/nft_range.c

···
 	priv->len = desc_from.len;
 	return 0;
 err2:
-	nft_data_uninit(&priv->data_to, desc_to.type);
+	nft_data_release(&priv->data_to, desc_to.type);
 err1:
-	nft_data_uninit(&priv->data_from, desc_from.type);
+	nft_data_release(&priv->data_from, desc_from.type);
 	return err;
 }