[PATCH 1/3] Introduce new phy_link and phy_instance abstraction

[Harald Welte]

Hi Holger,

On Fri, Feb 05, 2016 at 01:37:00PM +0100, Holger Freyther wrote:

> small memleak. plink being assigned twice.

thanks, fixed both now.

> > +	pcc->TrxId.byTrxId = pinst->u.octphy.trx_id;
> 
> ah nice, is there another way we could check that all primitives have
> all zero based indices set?

I'm not sure what you mean here?  AFAIK, the byTrxId can have arbitrary
values from the PHY side, we just need to use the same value as used in
the TRX-OPEN.req for all primitives belinging to that TRX.  However,
internally we use them with 0...n.

> okay ran out of time here. The octphy change looked good, mechanical,
> already fixing byTrxId assignment in case there will be more than one
> phy.

FYI: In the OCTPHY caes, we can actually have a single board
(OCTBTS3500) which has two DSPs.  To each of them we have pne PHY Link,
and in each of them we can have multiple Software TRX, which then map to
PHY instacnes on the OsmoBTS side.  Let's say you have two Soft TRX in
each DSP:
* in terms of gsm_bts_trx numbers, we have TRX 0 .. 3 to match our data
  model (and the way how the GSM specs deal with multiple TRX)
* in terms of byTrxId, we will have 0+1 on each of the PHY links

Regards,
	Harald
-- 
- Harald Welte <laforge at gnumonks.org>           http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
                                                  (ETSI EN 300 175-7 Ch. A6)