[PATCH] Support of multiple RSL connections for ABIS/ipaccess (BTS side)

[Andreas Eversberg]

Holger Hans Peter Freyther wrote:

hi holger,
> What does that it mean? What was the result of the test? What was
> the content of the test. "I tested it" is a sentence without any
> content. ;)
i setup a UmTRX with two transceivers and used the jolly/trx_rebased
branch to develop and test this patch. then i made calls on timeslots of
different TRX, to see if the correct RSL connection is used to handle
the channel. RSL connections on both TRX worked.
>> +			if (link->ofd->priv_nr >= E1INP_SIGN_RSL)
>> +				trx_nr = link->ofd->priv_nr - E1INP_SIGN_RSL;
>>     
>> +		e1i_ts = &link->line->ts[link->ofd->priv_nr-1];
>>     
> Where does this difference come from?
>   
the priv_nr is set to E1INP_SIGN_OML (1) or E1INP_SIGN_RSL+trx_nr (2..).
in order to retrieve the trx_nr from priv_nr, i substract
E1INP_SIGN_RSL. the priv_nr is also the index for selecting the ts
structure. because we do not have a physical E1 line, we can use ts[0]
also, so '1' is substracted.
> What are the alternatives to this patch? Can we copy the cfg.ipa fields
> into a per RSL line field and set the trx_id there before the first time
> we use it?
>   
i looked at the data structure. the ipaccess_unit structure, which is
linked to cfg.ipa.dev could be removed. instead we could provide
individual ipaccess_unit structure per RSL/OML link. when calling
e1inp_ipa_bts_rsl_connect(), we could provide a pointer to the
individual structure. this pointer must be stored inside ipa_client_conn
structure then.

alternatively we could provide an array of ipaccess_unit, instead of a
single structure and attach it to cfg.ipa.dev.

in both cases, the BTS code must care of filling the ipaccess_unit
structures with data for OML and each RSL link.

regards,

andreas