Eric Woudstra <[email protected]> wrote:
> This adds the capability to evaluate 802.1ad, QinQ, PPPoE and PPPoE-in-Q
> packets in the bridge filter chain.

Same comment as previous patch, this needs to explain the why, not the what.

nft_do_chain_bridge() passes all packets to the interpreter, so the
above statement is not correct either, you can already filter on all of
these packet types.  This exposes NFT_PKTINFO_L4PROTO etc, which is
different than what this commit message says.

I also vaguely remember I commented that this changes (breaks?) existing
behaviour for a rule like "tcp dport 22 accept" which may now match e.g.
a PPPoE packet.

Pablo, whats your take on this?  Do we need a new NFPROTO_BRIDGE
expression that can munge (populate) nft_pktinfo with the l4 data?

That would move this off to user policy (config) land.

(or extend nft_meta_bridge, doesn't absolutely require a brand new expression).

> +static inline int nft_set_bridge_pktinfo(struct nft_pktinfo *pkt,
> +                                      struct sk_buff *skb,
> +                                      const struct nf_hook_state *state,
> +                                      __be16 *proto)

This is only called from one .c file and unless this gets changed later
this should reside in that .c file (without inline keyword).

Reply via email to