Re: [PATCH net] net: openvswitch: fix possible memleak on create vport fails

[Jakub Kicinski <jakub.kicinski@netronome.com>]

On Tue, 17 Sep 2019 23:40:08 +0800, xiangxia.m.yue@gmail.com wrote:
> From: Tonghao Zhang <xiangxia.m.yue@gmail.com>
> 
> If we register a net device which name is not valid
> (dev_get_valid_name), register_netdevice will return err
> codes and will not run dev->priv_destructor. The memory
> will leak. This patch adds check in ovs_vport_free and
> set the vport NULL.
> 
> Fixes: 309b66970ee2 ("net: openvswitch: do not free vport if register_netdevice() is failed.")
> Cc: Taehee Yoo <ap420073@gmail.com>
> Signed-off-by: Tonghao Zhang <xiangxia.m.yue@gmail.com>

Thanks for the patch, I see what you're trying to do, but..

> diff --git a/net/openvswitch/vport-internal_dev.c b/net/openvswitch/vport-internal_dev.c
> index d2437b5..074c43f 100644
> --- a/net/openvswitch/vport-internal_dev.c
> +++ b/net/openvswitch/vport-internal_dev.c
> @@ -159,7 +159,6 @@ static struct vport *internal_dev_create(const struct vport_parms *parms)
>  	struct internal_dev *internal_dev;
>  	struct net_device *dev;
>  	int err;
> -	bool free_vport = true;
>  
>  	vport = ovs_vport_alloc(0, &ovs_internal_vport_ops, parms);
>  	if (IS_ERR(vport)) {
> @@ -190,10 +189,8 @@ static struct vport *internal_dev_create(const struct vport_parms *parms)
>  
>  	rtnl_lock();
>  	err = register_netdevice(vport->dev);
> -	if (err) {
> -		free_vport = false;
> +	if (err)
>  		goto error_unlock;
> -	}
>  
>  	dev_set_promiscuity(vport->dev, 1);
>  	rtnl_unlock();
> @@ -207,8 +204,7 @@ static struct vport *internal_dev_create(const struct vport_parms *parms)
>  error_free_netdev:
>  	free_netdev(dev);
>  error_free_vport:
> -	if (free_vport)
> -		ovs_vport_free(vport);
> +	ovs_vport_free(vport);
>  error:
>  	return ERR_PTR(err);
>  }
> diff --git a/net/openvswitch/vport.c b/net/openvswitch/vport.c
> index 3fc38d1..281259a 100644
> --- a/net/openvswitch/vport.c
> +++ b/net/openvswitch/vport.c
> @@ -157,11 +157,20 @@ struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
>   */
>  void ovs_vport_free(struct vport *vport)
>  {
> +	/* We should check whether vport is NULL.
> +	 * We may free it again, for example in internal_dev_create
> +	 * if register_netdevice fails, vport may have been freed via
> +	 * internal_dev_destructor.
> +	 */
> +	if (unlikely(!vport))
> +		return;
> +
>  	/* vport is freed from RCU callback or error path, Therefore
>  	 * it is safe to use raw dereference.
>  	 */
>  	kfree(rcu_dereference_raw(vport->upcall_portids));
>  	kfree(vport);
> +	vport = NULL;

vport here is a function argument, seems like setting it to NULL 
right before the function ends will do nothing. Should we rather 
set internal_dev->vport to NULL somehow? 

Perhaps someone more familiar with OvS can chime in and review..

>  }
>  EXPORT_SYMBOL_GPL(ovs_vport_free);
>