Re: [RFC 1/3] bpf/wireless: add wifimon program type

[Johannes Berg <johannes@sipsolutions.net>]

On Wed, 2017-04-12 at 11:19 -0400, David Miller wrote:
> 
> > Instead it may make more sense to just have a "wifi_info(skb,
> info)"
> > function you can call, e.g. with a structure that has various
> fields
> > and flags to see which you care about.
> 
> I would advise against this, let the parsing part use __sk_buff and
> therefore have maximum flexibility.  You really cannot predict the
> future, so in my opinion it might be unwise to constrain at this
> point.

So my point with the wifi_info() function to call from the BPF program
was just that putting something that's not already in struct sk_buff
into __sk_buff doesn't really make a lot of sense - we still have to
actually build it somewhere/somehow without knowing if it's going to be
needed by the program. We already have things like prog->cb_access and
prog->dst_needed now, but I'm not sure we want to blow that up much.

At the same time, technically I *do* have the information in skb->cb,
but the format thereof is something I really wouldn't want to let
trickle into the ABI. Therefore, I think if somebody needs something
like the bitrate, we should have a wifi_info() function that can return
the bitrate (among other things) without having to pre-build the data.
We can still cache it in mac80211 if multiple programs are there, dunno
if that makes sense.

Nevertheless, I think if I don't use __sk_buff as the program argument
then things get really messy, so I'll stick to that, and avoid adding
anything to it as much as possible.

johannes