On Wed, 2015-06-10 at 14:26 +0200, Matthias May wrote: > I'm not sure what exactly you mean that the handle_channel_custom needs > to loop through 5/10/20. > The freq_reg_info_regd sets the disabled flag on the channel at init. That's kinda my point - this isn't true. freg_req_info_regd() *doesn't* change any channel flags, it just returns some information, and through freq_reg_info() this is even exported to drivers. > This is not while trying to start operation on a channel, because the > channel is already disabled. > Or do you mean to actually have different sets of flags for the > different bandwidths? I mean basically what you did below, except handle it explicitly at the outer level. > I see that it's erroneously possible to run a 10/20MHz channel on 5175, > even though this shouldn't be allowed. > However setting the proper flags bw_flags should fix this. > > Expanding the patch with the patch below ensures that one can't start > operation on a channel which it's not allowed on. > Or is there a better way? I was thinking you should extend freq_reg_info_regd() to get the desired bandwidth, in which case you'd have to call it three times (5, 10, 20). However, you could also make the API so that freq_reg_info_regd() gets the *minimum* bandwidth [that the caller is willing to handle]. That way, you could set the argument to 20 MHz (for now) in freq_reg_info(), and to 5 MHz in the other callers that you're modifying (with the proposed additional patch) to actually handle the returned bandwidth lower than 20 MHz. This is really the crux of the problem - with your patch, freq_reg_info_regd() and thus freq_reg_info() can return something that doesn't fit the requested channel. You're changing the semantics of this function to no longer mean "return the rule that fits a 20MHz channel around the given center frequency" but to mean "return a rule around the given center frequency" - no matter how wide that rule can allow channels to be. If you change the function to pass in a "minimum required bandwidth" argument *and* change the checking as suggested with the combination, I think it'll be OK. johannes -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
