On Sun, 2009-01-18 at 07:28 -0800, Luis R. Rodriguez wrote: > Just so you know the approach I started taking for v3 of this series > was to apply save the orig_* parameters if REGDOM_SET_BY_DRIVER and > the request->wiphy == wiphy. This works OK except if you have "US" > as your regulatory domain and you have OLD_REG set cfg80211 will have > already have called for a new regulatory domain update for "US" so > this is -EALREADY. Some changes are then required to inform drivers > through the reg_notifier() of this. This is where I was at now. Yeah, I know. > > I think you're hampered by not being able to simply request a > > regdomain from crda outside the normal "apply regdomain" flow, > > Well I have a few approaches already which have worked but I like to > think we're looking for the most simple solution so its easier to > support this for other drivers as well. Right. > > maybe we > > should think about that instead? Then ath9k could just request a > > regdomain from crda before it even registers its wiphy? > > Are you suggesting we consider new API for drivers which want the > behaviour we want? That works for me too, but then again I'd be applying > the save orig_* stuff perhaps at the end of this new, say, > regulatory_hint_strict(). Well, I was more thinking to have a struct regdomain *cfg80211_ask_crda(const char *alpha2); but I realise that is very hard to implement with the call to userspace etc. So yeah, doing something like the _strict you propose seems more appropriate. Maybe call it something else, not _hint,and have it call out to userspace regardless (i.e. don't ever ignore that hint, as we might ignore other hints), say regulatory_apply(wiphy, alpha2)? Not sure, apply seems not quite right either. johannes
Attachment:
signature.asc
Description: This is a digitally signed message part