Re: [PATCH net-next v1 3/3] net/mlx5e: TC, Add support for VxLAN GBP encap/decap flows offload

[Alexander Lobakin <aleksander.lobakin@intel.com>]

From: Gavin Li <gavinl@nvidia.com>
Date: Wed, 15 Feb 2023 10:50:04 +0800

> 
> On 2/14/2023 11:26 PM, Alexander Lobakin wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> From: Gavin Li <gavinl@nvidia.com>
>> Date: Tue, 14 Feb 2023 15:41:37 +0200
>>
>>> Add HW offloading support for TC flows with VxLAN GBP encap/decap.

[...]

>>> @@ -96,6 +99,70 @@ static int mlx5e_gen_ip_tunnel_header_vxlan(char
>>> buf[],
>>>        udp->dest = tun_key->tp_dst;
>>>        vxh->vx_flags = VXLAN_HF_VNI;
>>>        vxh->vx_vni = vxlan_vni_field(tun_id);
>>> +     if (tun_key->tun_flags & TUNNEL_VXLAN_OPT) {
>>> +             md = ip_tunnel_info_opts((struct ip_tunnel_info
>>> *)e->tun_info);
>>> +             vxlan_build_gbp_hdr(vxh, tun_key->tun_flags,
>>> +                                 (struct vxlan_metadata *)md);
>> Maybe constify both ip_tunnel_info_opts() and vxlan_build_gbp_hdr()
>> arguments instead of working around by casting away?
> The reason to cast it is a WA to use ip_tunnel_info_opts which takes
> non-const arg while>
> e->tun_info here is a const one.

That's what I'm asking - why not make ip_tunnel_info_opts() and
vxlan_build_gbp_hdr() take const tun_info?

> 
>>
>>> +     }
>>> +
>>> +     return 0;
>>> +}
[...]

Thanks,
Olek