netdev - Re: [PATCH net-next 5/7] net: add dst

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Thu, 11 Feb 2016 12:48:38 +0100
From:	Jiri Benc <jbenc@...hat.com>
To:	Paolo Abeni <pabeni@...hat.com>
Cc:	netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
	Thomas Graf <tgraf@...g.ch>,
	Pravin B Shelar <pshelar@...ira.com>,
	Hannes Frederic Sowa <hannes@...essinduktion.org>,
	Tom Herbert <therbert@...gle.com>
Subject: Re: [PATCH net-next 5/7] net: add dst_cache to ovs vxlan lwtunnel

On Thu, 11 Feb 2016 11:12:01 +0100, Paolo Abeni wrote:
> In case of UDP traffic with datagram length
> below MTU this give about 2% performance increase

The performance increase is not that great probably because of the
addition of the pointer to ip_tunnel_info, making it even fatter than
it is now.

> diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
> index d1bd4a4..f181186 100644
> --- a/net/openvswitch/flow_netlink.c
> +++ b/net/openvswitch/flow_netlink.c
> @@ -1698,6 +1698,7 @@ static void ovs_nla_free_set_action(const struct nlattr *a)
>  	case OVS_KEY_ATTR_TUNNEL_INFO:
>  		ovs_tun = nla_data(ovs_key);
>  		dst_release((struct dst_entry *)ovs_tun->tun_dst);
> +		dst_cache_destroy(&ovs_tun->dst_cache);

We need a helper function for this, operating on ovs_tunnel_info.

>  		break;
>  	}
>  }
> @@ -1928,6 +1929,7 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
>  {
>  	struct sw_flow_match match;
>  	struct sw_flow_key key;
> +	struct dst_cache dst_cache;
>  	struct metadata_dst *tun_dst;
>  	struct ip_tunnel_info *tun_info;
>  	struct ovs_tunnel_info *ovs_tun;
> @@ -1959,15 +1961,24 @@ static int validate_and_copy_set_tun(const struct nlattr *attr,
>  	if (!tun_dst)
>  		return -ENOMEM;
>  
> +	err = dst_cache_init(&dst_cache, GFP_KERNEL);
> +	if (err) {
> +		dst_release((struct dst_entry *)tun_dst);
> +		return err;
> +	}
> +
>  	a = __add_action(sfa, OVS_KEY_ATTR_TUNNEL_INFO, NULL,
>  			 sizeof(*ovs_tun), log);
>  	if (IS_ERR(a)) {
>  		dst_release((struct dst_entry *)tun_dst);
> +		dst_cache_destroy(&dst_cache);

Make the local variable be of ovs_tunnel_info type and use the helper
function here.

>  		return PTR_ERR(a);
>  	}
>  
>  	ovs_tun = nla_data(a);
>  	ovs_tun->tun_dst = tun_dst;
> +	ovs_tun->dst_cache = dst_cache;

Why are you copying the data here? The cache should be initialized in
place in ovs_tun.

> +	tun_dst->u.tun_info.dst_cache = &ovs_tun->dst_cache;

The absence of reference counting here will lead to use after free when
processing a packet referencing tun_dst while the corresponding
dst_cache memory is freed on flow deletion. Note that tun_dst is
reference counted (see execute_set_action).

 Jiri

-- 
Jiri Benc