On Thu, 2007-10-18 at 18:48 +0200, Johannes Berg wrote: > Hmm. And here's a thought: don't we need all this information within > mac80211 as well? Would it maybe make sense to embed it into the sta > sdata structure and have the 'changed_map' not be in the structure but > as a second parameter to the callback function? Then we could always use > a pointer to the embedded structure that keeps track of this information > and build the changed value dynamically. This saves having to initialise > the structure all the time when calling the function and makes sure that > even the unchanged parameters are always valid should the driver need > them. [also see at the very end of this mail] Re-reading my own mail :) Don't we need an if_id parameter to the callback so if the driver supports multiple virtual STA interfaces it can make an effort to support the settings of both BSSes? I'd be tempted to do it in mac80211 but I guess that different MAC designs have different levels of configurability when multiple addresses are programmed so it seems the only sane place to do it would be the driver since mac80211 cannot know which settings are only supported once and where different settings can be supported by the MAC. johannes
Attachment:
signature.asc
Description: This is a digitally signed message part
