The sgsn uses other data structs so don't require them inside libiu.
Instead keep a private list of ue contexts and iterate through that.
This commit reverts the libui changes of commit
d03faa4bacd4d2a8b9155faf5219a948b73f481c
In the IU case the RNC and ggsn communicate directly on the user plane.
Since the IP address of the RNC is not known in our case (it sits behind
the hnbgw) we need to update the PDP context with the new IP address
after receiving the RAB assignment response (which includes the IP
address).
The RAB assignment response includes the gtp teid that we sent along in
the assignment request. Retrieve the correct pdp context from there and
activate it.
Bit 4 is reserved in 3GPP TS 04.08 so exclude it from the type.
In 3GPP TS 24.008 it indicates if a follow-on request is pending by the
MS, but only in Iu mode. According to the spec it is not required to
react to that request with a follow-on proceed so this field can be
ignored for now.
See 3GPP TS 24.008 Ch. 4.4:
"Unless it has specific permission from the network (follow-on proceed)
the mobile station side should await the release of the RR connection
used for a MM specific procedure before a new MM specific procedure or
MM connection establishment is started."
as well as Ch. 4.4.4.6:
"If the network wishes to prolong the RR connection to allow the mobile
station to initiate MM connection establishment (for example if the
mobile station has indicated in the LOCATION UPDATING REQUEST that it
has a follow-on request pending) the network shall send "follow on
proceed" in the LOCATION UPDATING ACCEPT and start timer T3255."
NITB is already gone from this branch. We may even resurrect it, but I prefer
to name the realms 'libmsc' and 'libbsc', hence the nitb graph is the BSC graph
now. Also add a libmsc graph.
The aim is to allow osmo-cscn to pass its own root talloc context and
global gsm_network struct instance cleanly. This may stir up some old and
dusty globals, but I hope it's for the better, since not all is a BSC.
To ensure that a global gsm_network pointer for the bsc_vty is set, have it as
argument to bsc_vty_init(). The vty configuration commands are added only after
bsc_vty_init(), which are needed to configure the network struct. So split up
the bsc_bootstrap_network() function into bsc_network_init() to allocate a
gsm_struct, and bsc_network_configure() to read the config file once the vty
commands are in place. In this way, no global bsc_gsmnet pointer is needed for
the bsc vty. The atomic super glue is dissolved and osmo-cscn will be allowed
to have a different name for it.
Admitted, it's still called the bsc_vty, but a split thereof is probably coming
soon, because the CSCN doesn't want any of the BSC nor BTS specific vty
commands.
The diff between this and master will probably need a lot of review and fixes.
The current state does compile, but I expect pretty much everything to be
broken now. Future development will reinstate proper functionality piecemeal.
The first goal is to get basic signalling to work, then SMS. The voice control
(RTP) is completely disabled now (see "#if BEFORE_MSCSPLIT") and will be fixed
last AFAICT.
Pubish msc_compl_l3() decl in new file libmsc/msc_api.h (but see comment).
Call msc_compl_l3() when establishing a subscriber connection for IuCS.
Remove bts from subscr_conn_allocate_iu() signature, use network, link_id and
conn_id instead.
Move subscr_conn_allocate_iu() to the top of the file, because it semantically
belongs before subscr_conn_lookup_iu().
Add libiu to contain the parts used by both Iu-CS (in osmo-cscn) and Iu-PS (in
gprs) into libiu. It's rather thin and may make sense to move to osmo-iuh
altogether, eventually.
iu.c is half moved to libiu/, and half to osmo-cscn/iu_cs.c.
