On 08/03/2024 03:04, Andrew Lunn wrote:

quoted

+static inline bool ovpn_peer_hold(struct ovpn_peer *peer)
+{
+	return kref_get_unless_zero(&peer->refcount);
+}
+
+static inline void ovpn_peer_put(struct ovpn_peer *peer)
+{
+	kref_put(&peer->refcount, ovpn_peer_release_kref);
+}

It is reasonably normal in the kernel to use _get() which takes a
reference on something and _put() to release it.

I think I got inspired by dev_hold(), but I agree that _get() is more 
appropriate.

Will change that.

quoted

+struct ovpn_peer *ovpn_peer_lookup_transp_addr(struct ovpn_struct *ovpn, struct sk_buff *skb);
+struct ovpn_peer *ovpn_peer_lookup_by_dst(struct ovpn_struct *ovpn, struct sk_buff *skb);
+struct ovpn_peer *ovpn_peer_lookup_by_src(struct ovpn_struct *ovpn, struct sk_buff *skb);
+struct ovpn_peer *ovpn_peer_lookup_id(struct ovpn_struct *ovpn, u32 peer_id);

All these look to take a reference on the peer. So maybe replace
lookup by get? It should then be easier to check there is a matching
put to every get.

Alright, will do!

Regards,

-- 
Antonio Quartulli
OpenVPN Inc.