On 07/21/2016 09:27 AM, Jamal Hadi Salim wrote: [...]
Forgot about csum which would work with pedit - didnt quiet parse what you are saying above though: does changing MAC address require changing to CHECKSUM_NONE? If yes, then seems like i need to send a patch for act_ife as well to make in the spirit of mirred?
Hmm, seems fine actually up to ETH_HLEN. In cases like vlan or mpls modifications there's certainly a correction, but when going into rx path this is pulled out already of the calculation afaik (see f.e. dev_forward_skb()). So when mirred pushes this back in via skb_push_rcsum() and dev_forward_skb() back out, it should be okay if I see this correctly. But don't you still something like need skb_try_make_writable()?
