Re: [RFC bluetooth-next 09/15] ieee802154: add support for get tx powers

[Alexander Aring <alex.aring@gmail.com>]

Hi Phoebe,

On Mon, Apr 27, 2015 at 01:49:32PM +0200, Phoebe Buckheister wrote:
> On Thu, 23 Apr 2015 18:47:48 +0200
> Alexander Aring <alex.aring@gmail.com> wrote:
> 
...
> > +	int		(*get_tx_powers)(struct ieee802154_hw
> > *hw, s8 *dbm,
> > +					 u32 idx);
> 
> I suggest you use mBm instead of dBm here, while you're at it anyway.
> Especially since one of your later patches will add incorrect values
> because dBm isn't accurate enough for what the transceiver support
> (rf230, TX power +3.4dBm). This would also make the value-2 return you
> have right now unnecessary.
> 

The 802.15.4 standard describes in the phy pib for tx power:

attr: phyTXPower
type: signed integer
range: -
describtion: The transmit power of the device in dBm.

I mentioned in my previous mail that this doesn't repsonse the reality
because there are transceivers outside which supports tx power in dbm
which are not signed integer, like +3.4 dbm.

Nevertheless I would say forget that what the PIB says, we do what the
reality does now. (In my previous mail I was another opinion because the
current type is s8). I think we should not do that now, because we get
trouble later, this is what you mentioned now and I agree.


Plan to implement the mbm:

1. change current tx power to S32 instead s8.
   I gave it a try phy dump with iwpan still works, (not ends in a
   critical error), so I will change it.

2. the kernelside use mbm everywhere the userspace side kann convert the
   dbm values to mbm then with a calculation "dbm = mbm / 100" and
   "mbm = dbm * 100".

   Some helper marcos like:
   #define DBM_TO_MBM(gain) ((int)(((float)gain) * 100))
   #define MBM_TO_DBM(gain) ((float)(gain) / 100)

Examples
   MBM    | DBM
   +230   | +2.3
   -280   | -2.8
   ...    | ....

Thanks, for the review.

- Alex