netdev - Re: [PATCH bpf,v5 3/4] bpf: fix bpf socket lookup from tc/xdp to respect socket VRF bindings

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] [day] [month] [year] [list]

Message-ID: <1a36a60c-ec8f-9927-a932-f4fda0a52af3@iogearbox.net>
Date: Wed, 21 Jun 2023 00:30:57 +0200
From: Daniel Borkmann <daniel@...earbox.net>
To: Gilad Sever <gilad9366@...il.com>, dsahern@...nel.org,
 martin.lau@...ux.dev, john.fastabend@...il.com, ast@...nel.org,
 andrii@...nel.org, song@...nel.org, yhs@...com, kpsingh@...nel.org,
 sdf@...gle.com, haoluo@...gle.com, jolsa@...nel.org, davem@...emloft.net,
 edumazet@...gle.com, kuba@...nel.org, pabeni@...hat.com, mykolal@...com,
 shuah@...nel.org, hawk@...nel.org, joe@...d.net.nz
Cc: eyal.birger@...il.com, shmulik.ladkani@...il.com, bpf@...r.kernel.org,
 netdev@...r.kernel.org, linux-kselftest@...r.kernel.org
Subject: Re: [PATCH bpf,v5 3/4] bpf: fix bpf socket lookup from tc/xdp to
 respect socket VRF bindings

On 6/8/23 1:41 PM, Gilad Sever wrote:
> When calling bpf_sk_lookup_tcp(), bpf_sk_lookup_udp() or
> bpf_skc_lookup_tcp() from tc/xdp ingress, VRF socket bindings aren't
> respoected, i.e. unbound sockets are returned, and bound sockets aren't
> found.
> 
> VRF binding is determined by the sdif argument to sk_lookup(), however
> when called from tc the IP SKB control block isn't initialized and thus
> inet{,6}_sdif() always returns 0.
> 
> Fix by calculating sdif for the tc/xdp flows by observing the device's
> l3 enslaved state.
> 
> The cg/sk_skb hooking points which are expected to support
> inet{,6}_sdif() pass sdif=-1 which makes __bpf_skc_lookup() use the
> existing logic.
> 
> Fixes: 6acc9b432e67 ("bpf: Add helper to retrieve socket in BPF")
> Acked-by: Stanislav Fomichev <sdf@...gle.com>
> Reviewed-by: Shmulik Ladkani <shmulik.ladkani@...il.com>
> Reviewed-by: Eyal Birger <eyal.birger@...il.com>
> Signed-off-by: Gilad Sever <gilad9366@...il.com>

Overall this series looks good to go, just small nits which would be great
to still address.

> ---
> v5: Use reverse xmas tree indentation
> 
> v4: Move dev_sdif() to include/linux/netdevice.h as suggested by Stanislav Fomichev
> 
> v3: Rename bpf_l2_sdif() to dev_sdif() as suggested by Stanislav Fomichev
> ---
>   include/linux/netdevice.h |  9 +++++++
>   net/core/filter.c         | 54 ++++++++++++++++++++++++---------------
>   2 files changed, 42 insertions(+), 21 deletions(-)
> 
> diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
> index c2f0c6002a84..db1bfca6b8b4 100644
> --- a/include/linux/netdevice.h
> +++ b/include/linux/netdevice.h
> @@ -5093,6 +5093,15 @@ static inline bool netif_is_l3_slave(const struct net_device *dev)
>   	return dev->priv_flags & IFF_L3MDEV_SLAVE;
>   }
>   
> +static inline int dev_sdif(const struct net_device *dev)
> +{
> +#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)

Why IS_ENABLED? config NET_L3_MASTER_DEV says bool, so #ifdef CONFIG_NET_L3_MASTER_DEV
should suffice?

> +	if (netif_is_l3_slave(dev))
> +		return dev->ifindex;
> +#endif
> +	return 0;
> +}
> +
>   static inline bool netif_is_bridge_master(const struct net_device *dev)
>   {
>   	return dev->priv_flags & IFF_EBRIDGE;
[...]

>   static const struct bpf_func_proto bpf_tc_sk_lookup_tcp_proto = {
> @@ -6778,10 +6782,11 @@ BPF_CALL_5(bpf_tc_sk_lookup_udp, struct sk_buff *, skb,
>   {
>   	struct net *caller_net = dev_net(skb->dev);
>   	int ifindex = skb->dev->ifindex;
> +	int sdif = dev_sdif(skb->dev);
>   
>   	return (unsigned long)__bpf_sk_lookup(skb, tuple, len, caller_net,
>   					      ifindex, IPPROTO_UDP, netns_id,
> -					      flags);
> +					      flags, sdif);
>   }
>   
>   static const struct bpf_func_proto bpf_tc_sk_lookup_udp_proto = {
> @@ -6814,11 +6819,13 @@ BPF_CALL_5(bpf_xdp_sk_lookup_udp, struct xdp_buff *, ctx,
>   	   struct bpf_sock_tuple *, tuple, u32, len, u32, netns_id, u64, flags)
>   {
>   	struct net *caller_net = dev_net(ctx->rxq->dev);
> -	int ifindex = ctx->rxq->dev->ifindex;
> +	struct net_device *dev = ctx->rxq->dev;

Why not doing this before the struct net *caller_net and also use it there
for the dev_net()? Same in other XDP places. It would be nice to also do the
same for the tc helpers.

> +	int ifindex = dev->ifindex;
> +	int sdif = dev_sdif(dev);
>   
>   	return (unsigned long)__bpf_sk_lookup(NULL, tuple, len, caller_net,
>   					      ifindex, IPPROTO_UDP, netns_id,
> -					      flags);
> +					      flags, sdif);
>   }
>   
>   static const struct bpf_func_proto bpf_xdp_sk_lookup_udp_proto = {
> @@ -6837,11 +6844,13 @@ BPF_CALL_5(bpf_xdp_skc_lookup_tcp, struct xdp_buff *, ctx,
>   	   struct bpf_sock_tuple *, tuple, u32, len, u32, netns_id, u64, flags)
>   {
>   	struct net *caller_net = dev_net(ctx->rxq->dev);
> -	int ifindex = ctx->rxq->dev->ifindex;
> +	struct net_device *dev = ctx->rxq->dev;
> +	int ifindex = dev->ifindex;
> +	int sdif = dev_sdif(dev);
>   
>   	return (unsigned long)__bpf_skc_lookup(NULL, tuple, len, caller_net,
>   					       ifindex, IPPROTO_TCP, netns_id,
> -					       flags);
> +					       flags, sdif);
>   }
>   
>   static const struct bpf_func_proto bpf_xdp_skc_lookup_tcp_proto = {