Re: [PATCH 2/2] mac80211: make listen_interval be configurable by low level driver

[Kalle Valo <kalle.valo@nokia.com>]

Tomas Winkler <tomasw@gmail.com> writes:

>>
>> This stuff is very much hardware dependent, I don't deny that, but
>> still I would assume that the basic principles are almost the same. We
>> just need to come up with a good interface for the drivers. Sure, it's
>> more work now, but if we don't do it, PSM support will eventually get
>> really messy because all drivers do it somehow differently.
>>
> This patch only sets value of listen interval from the driver I just
> think this is exactly what you are suggesting.

No, I don't suggest that. I understood that for setting the
listen_interval you are after something like this:

         sysfs?                 register_hw()
user   ---------->   driver    -------------->  mac80211

(Please correct me if I misunderstood this.)

But I think that the flow should be like this:

       WE/cfg80211                config()
user   ---------->   mac80211  -------------->  driver

So that mac80211 would be one making the choice for the listen value
based on user input. But what the user input would be, that's an
another question.

> This is only required for sending correct value of listen interval
> to AP in association. Anything else is handled in the driver or
> firmware anyway.

Yes, driver will pass the information to the firmware. But I still
think mac80211 should make the decision, and the driver should be only
forwarding mac80211 decisions to the firmware.

> This patch should go in and we can reconsider it when other HW
> driver will implement PS so we can see the bigger picture.

I guess it's a good compromise for now, but as long as we change it
later.

-- 
Kalle Valo