Re: [PATCH RFC v4 bpf-next 03/11] xdp: Add xdp_txq_info to xdp_buff

[Jesper Dangaard Brouer <brouer@redhat.com>]

On Fri, 28 Feb 2020 11:07:23 +0100
Toke Høiland-Jørgensen <toke@redhat.com> wrote:

> David Ahern <dahern@digitalocean.com> writes:
> 
> > On 2/27/20 1:00 AM, Jesper Dangaard Brouer wrote:  
> >>> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >>> index 7850f8683b81..5e3f8aefad41 100644
> >>> --- a/include/uapi/linux/bpf.h
> >>> +++ b/include/uapi/linux/bpf.h
> >>> @@ -3334,8 +3334,10 @@ struct xdp_md {
> >>>  	__u32 data;
> >>>  	__u32 data_end;
> >>>  	__u32 data_meta;
> >>> -	/* Below access go through struct xdp_rxq_info */
> >>> -	__u32 ingress_ifindex; /* rxq->dev->ifindex */
> >>> +	union {
> >>> +		__u32 ingress_ifindex; /* rxq->dev->ifindex */
> >>> +		__u32 egress_ifindex;  /* txq->dev->ifindex */
> >>> +	};  
> >> 
> >> Are we sure it is wise to "union share" (struct) xdp_md as the
> >> XDP-context in the XDP programs, with different expected_attach_type?
> >> As this allows the XDP-programmer to code an EGRESS program that access
> >> ctx->ingress_ifindex, this will under the hood be translated to
> >> ctx->egress_ifindex, because from the compilers-PoV this will just be an
> >> offset.
> >> 
> >> We are setting up the XDP-programmer for a long debugging session, as
> >> she will be expecting to read 'ingress_ifindex', but will be getting
> >> 'egress_ifindex'.  (As the compiler cannot warn her, and it is also
> >> correct seen from the verifier).  
> >
> > It both cases it means the device handling the packet. ingress_ifindex
> > == device handling the Rx, egress_ifindex == device handling the Tx.
> > Really, it is syntactic sugar for program writers. It would have been
> > better had xdp_md only called it ifindex from the beginning.  
> 
> Telling users that they are doing it wrong is not going to make their
> debugging session any less frustrating :)
> 
> If we keep rx_ifindex a separate field we can unambiguously reject a TX
> program that tries to access it, *and* we keep the option of allowing
> access to it later if it does turn out to be useful. IMO that is worth
> the four extra bytes.

I agree. We need unambiguously to help the program writer.

This is the wrong kind of 'syntactic sugar'.  If you want a straight
'ifindex', that translates to the running_ifindex, when you need to add
a new member 'ifindex' that does this rewriting based on attach type.

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer