Re: [PATCH V9 2/2] cfg80211/nl80211: Enable drivers to implement MAC address based ACL

[Johannes Berg <johannes@sipsolutions.net>]

On Fri, 2013-01-18 at 11:18 +0530, Vasanthakumar Thiagarajan wrote:

> + * @NL80211_ATTR_ACL_POLICY: ACL policy, see &enum nl80211_acl_policy_attr.
> + *
> + * @NL80211_ATTR_MAC_ADDRS: Array of nested MAC addresses, used for
> + *	MAC ACL.

Ok so one last (not a promise ;-) ) thing ...

You basically completely back-pedalled from wanting to support white &
blacklist at the same time to no longer even allowing that in the API.
Should I really merge this? Who says you're not going to return in a
couple of weeks and want that then, making the API very strange?

Right now, you have it structured like this:

POLICY=u8 value, ADDRS=[1=addr1, 2=addr2, ...]

(= indicates attribute type/value, [] indicates nesting)

I think it might be smarter to do

ACL=[TYPE=addr1|addr2|..., TYPE=addr3|addr4|...]

where "|" just indicates concatenation. I don't really strictly see a
reason to have a new nested attribute for each MAC address, since they
are each just 6 bytes long, for all I care they could be concatenated.
But if you wanted nesting, that would make it look like this:

ACL[TYPE=[1=addr1, 2=addr2, ...], TYPE=[1=addr3, 2=addr4, ...]]

which is acceptable too, just requires more code to parse.


Anyway, if you think that the dual list approach is dead indefinitely, I
guess I can apply it. I'm just not sure I believe that ;-)

johannes