Re: [PATCH ipsec-next 4/6] xfrm: add TX datapath support for IPsec full offload mode

[Leon Romanovsky <leon@kernel.org>]

On Fri, May 13, 2022 at 04:56:58PM +0200, Steffen Klassert wrote:
> On Tue, May 10, 2022 at 01:36:55PM +0300, Leon Romanovsky wrote:
> > From: Leon Romanovsky <leonro@nvidia.com>
> > 
> > In IPsec full mode, the device is going to encrypt and encapsulate
> > packets that are associated with offloaded policy. After successful
> > policy lookup to indicate if packets should be offloaded or not,
> > the stack forwards packets to the device to do the magic.
> > 
> > Signed-off-by: Raed Salem <raeds@nvidia.com>
> > Signed-off-by: Huy Nguyen <huyn@nvidia.com>
> > Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
> > ---
> >  net/xfrm/xfrm_output.c | 19 +++++++++++++++++++
> >  1 file changed, 19 insertions(+)
> > 
> > diff --git a/net/xfrm/xfrm_output.c b/net/xfrm/xfrm_output.c
> > index d4935b3b9983..2599f3dbac08 100644
> > --- a/net/xfrm/xfrm_output.c
> > +++ b/net/xfrm/xfrm_output.c
> > @@ -718,6 +718,25 @@ int xfrm_output(struct sock *sk, struct sk_buff *skb)
> >  		break;
> >  	}
> >  
> > +	if (x->xso.type == XFRM_DEV_OFFLOAD_FULL) {
> > +		struct dst_entry *dst = skb_dst_pop(skb);
> > +
> > +		if (!dst) {
> > +			XFRM_INC_STATS(net, LINUX_MIB_XFRMOUTERROR);
> > +			return -EHOSTUNREACH;
> > +		}
> > +
> > +		skb_dst_set(skb, dst);
> > +		err = skb_dst(skb)->ops->local_out(net, skb->sk, skb);
> > +		if (unlikely(err != 1))
> > +			return err;
> > +
> > +		if (!skb_dst(skb)->xfrm)
> > +			return dst_output(net, skb->sk, skb);
> > +
> > +		return 0;
> > +	}
> > +
> 
> How do we know that we send the packet really to a device that
> supports this type of offload? For crypto offload, we check that
> in xfrm_dev_offload_ok() and I think something similar is required
> here too.

I think that function is needed to make sure that we will have SW
fallback. It is not needed in full offload, anything that is not
supported/wrong should be dropped by HW.

> 
> Also, the offload type still requires software policies and states.
> What if a device comes up that can do a real full offload, i.e.
> in a way that the kernel acts just as a stub layer between IKE
> and the device. Are we going to create XFRM_DEV_OFFLOAD_FULL_2
> then? We need to make sure that this case cann be supported with
> the new API too.

Yes, I think that it is supported by this API.

From user perspective, all flavours of full offload are the same, the
difference is in-kernel API, where we will be able differentiate with
some sort of features flag.

Thanks