Re: [PATCH RFC bpf-next 1/2] bpf_redirect_neigh: Support supplying the nexthop as a helper parameter

[Daniel Borkmann <daniel@iogearbox.net>]

On 10/15/20 9:34 PM, Toke Høiland-Jørgensen wrote:
> David Ahern <dsahern@gmail.com> writes:
>> On 10/15/20 9:46 AM, Toke Høiland-Jørgensen wrote:
>>> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>>> index bf5a99d803e4..980cc1363be8 100644
>>> --- a/include/uapi/linux/bpf.h
>>> +++ b/include/uapi/linux/bpf.h
>>> @@ -3677,15 +3677,19 @@ union bpf_attr {
>>>    * 	Return
>>>    * 		The id is returned or 0 in case the id could not be retrieved.
>>>    *
>>> - * long bpf_redirect_neigh(u32 ifindex, u64 flags)
>>> + * long bpf_redirect_neigh(u32 ifindex, struct bpf_redir_neigh *params, int plen, u64 flags)
>>
>> why not fold ifindex into params? with params and plen this should be
>> extensible later if needed.
> 
> Figured this way would make it easier to run *without* the params (like
> in the existing examples). But don't feel strongly about it, let's see
> what Daniel thinks.

My preference is what Toke has here, this simplifies use by just being able to
call bpf_redirect_neigh(ifindex, NULL, 0, 0) when just single external facing
device is used.

>> A couple of nits below that caught me eye.
> 
> Thanks, will fix; the kernel bot also found a sparse warning, so I guess
> I need to respin anyway (but waiting for Daniel's comments and/or
> instructions on what tree to properly submit this to).

Given API change, lets do bpf. (Will review the rest later today.)