This change reproduces a segfault caused by the recent regression [1].
The IUT tries to bind on BSSGP host:port that it's already bound to.
[1] Ib389925cf5c9f18951af6242c31ea70476218e9a
Change-Id: I4c3bc883d795e5d1ee5ab175ac03684924692a7c
Related: SYS#5108
The BVC-RESET / BVC-RESEt-ACK follow a set of rules:
* Signaling BVCI=0 never has a CellId in BVC-RESET nor BVC-RESET-ACK
* Any BVC-RESET or BVC-RESET ack in BSS->SGSN direction must have CellID
* Any BVC-RESET or BVC-RESET ack in SGSN->BSS direction must NOT have CellID
Let's adjust our test expectations accordingly.
This will break tests against "latest", but the amount of work-arounds
needed in this code outweighs the benefit.
Change-Id: Ic8a83f5214c372faa15178dd9b54364e7d2a60cb
The existing BSSGP_Emulation is built around the assumption that every
instance of BSSGP_Emulation only servers one signaling-BVC and one
PTP-BVC. While this is true for osmo-pcu at this point (BTS-colocated
PCU), other BSS/PCU implementations differ.
In general, there can always be any number of PTP BVC (one per cell)
next to the signaling BVC (one per BSS). Let's represent this in
BSSGP_Emulation so we can create more comprehensive tests.
Change-Id: I7e30b4c4e188518a574e082962fba457b3a97ce3
This is required by the spec, and implemented libosmocore since
Change-Id Ie87820537d6d616da4fd4bbf73eab06e28fda5e1
So let's change our test expectations. Meanwhile, introduce
mp_tolerate_bvc_reset_cellid for working around the bug in 'latest'.
Change-Id: Icebee25b53fef623db6ae91ca0d943e70a3c86b7
This adds a NS_Provider_FR which interfaces between FrameRelay_Emulation
and NS_Emulation. Include support for it only if enabled at compile
time to avoid pulling in a dependency on the FrameRelay stack for every
user of NS_Emulation.
Change-Id: I42ca811d23e383e362d2527c8ff2c916a62a5b42
NSConfiguration currently contains parameters relevant only for IP
based transport. Move IP/UDP parameters into a sub-structure in
anticipation of Frame Relay support.
Change-Id: I6904520d8c2f546327029777d68b1907611a8cf5
Both osmo-bts and osmo-pcu are switching to PCUIFv10 soon, so let's
use the new version by default. For older (latest) IUT versions
not supporting PCUIFv10, one would need to override this module
parameter in {BTS,PCU}_Tests.cfg.
Change-Id: I9350c4a54434c3d46ce9424f382ca0057e58d053
Related: SYS#4868, SYS#4915
This change will facilitate adding test cases for multi-slot TBF
allocation across several TS and/or TRX instances of a BTS.
Change-Id: I9bff9e912bf73ad6259946a6ea5b08f3e2f728c8
Related: OS#4756
The dependency of PCUIF_Types creates also a dependency on
Replace the PCU_AddrType by an unix like address family defined
in the Osmocom_Types to reduce the dependency.
Change-Id: I0b4fda96accef401ffc009010f9f5621583fd6dd
NS_Emulation_CT:NSCP used to be a NS_CODEC_PT, which is a translation
port on top of the IPL4asp. Such ports need to be mapped to a system
port at start-up.
However, in I4d0b7ad0ed9447a038dd3eeee2b975146d10fba0 we introduced
a new underlying component called NS_Provider. Hence, NSCP is actually
connected to that underlying component, and no longer mapped.
The corresponding change in pcu/SGSN_Components.ttcn and sgsn/SGSN_Tests.ttcn
has unfortunately been missed in the above-mentioned change. Fix that.
Change-Id: I4c085210e6021e38a38ebc052ec3d9b345638cd2
This is a step to prepare the NS_Emulation for operating on top of
Frame Relay, not just UDP/IP. We replace the NS_CodecPort (mapping to
a IPL4asp) with a newly introduced NS_Provider_CT, specifically a
NS_Provider_IPL4_CT.
This change removes any IP specific bits from NS_Emulation and moves
it into the newly-created NS_ProvideR_IPL4.
Change-Id: I4d0b7ad0ed9447a038dd3eeee2b975146d10fba0
RAW_NS uses module parameter from SGSN_Components. To decouple
RAW_NS from SGSN_Component pass the cell_id via a function argument.
Change-Id: I91d9db85519675054aaab83c85fac43e67391f33
RAW_NS uses module parameter from SGSN_Components. To decouple
RAW_NS from SGSN_Component pass the NSConfiguration via f_init_ns_codec()
Change-Id: Ida8b8a6af815dc11b2ff4c65e19cc5ec25f18ae2
My initial assumption was that we can skip redundant '0'B bits or
even '00'O octets in the Mobile Allocation IE, and thus reduce
the overall size of this element. Unfortunately, this is wrong.
3GPP TS 44.018, section 10.5.2.21 clearly states that the Mobile
Allocation IE contains a bit-string of size NF, where NF is the
number of frequencies in the cell allocation. If NF % 8 != 0,
then '0'B padding bits must be appended to make it octet-aligned.
In other words, if the cell allocation contains let's say 13
frequencies, but a hopping timeslot makes use of only a small
fraction of it (e.g. 4 first channels), we would still need to
transmit at least 13 bits (+padding), including all redundant
bits and octets.
In RLC/MAC frames though it's not required to make the bit-string
octet aligned, so we need to send exactly NF bits without padding.
In order to achieve that:
a) add MA length field to INFO.ind (record PCUIF_InfoTrxTs);
b) ajust the existing test cases to use this field.
It's safe to merge this change as the related patches have not
been merged to osmo-pcu and osmo-bts yet.
Change-Id: I2709673c90a0cd7d76de9db8b8ad82ac59ca84a0
Related: SYS#4868, OS#4547
The ALIVE PDU does not have to be sent before an UNBLOCK message.
Ignore UNBLOCK messages in cases a single ALIVE is expected.
According to TS 48.016 v5.2.0, 7.4:
"Upon successful completion of an NS-VC reset procedure,
a BSS (or SGSN) shall start timer Tns-test, then [..]"
The ALIVE should arive Tns-test after the reset procedure (typ. 1s-60s).
Change-Id: I11d77b7477981998082967e5123b61636af2b980
osmo-pcu.git 0052051c07af63da98137c9f8e3b3119642eb587 introduced a bug
(fixed in 1d68cdff928f32941aff36c70e4543203c283d15) where no MS could
GMM attach to the network, but no test showed the issue.
This couple test verifies both correct behavior and also ensures wrong
IMSI is detected and reported.
Related: OS#4729
Change-Id: I5072d80b7ed9945a6083cdf3254efb8b8f119c54
This change introduces new version 10 specific extensions, in
particular: the frequency hopping parameters of each timeslot.
These parameters are used to compose Channel Description IE
in the packet resource assignment messages.
In order to maintain backwards compatibility with version 9 of
the PCUIF, and thus to still be able to run test cases against
the latest release of osmo-pcu, I kept the old parts of the
INFO.ind and gruoped them together with the new records
into union 'PCUIF_InfoTrxs'.
During decoding, the content of this union is resolved by the
TITAN's RAW codec itself, depending on value of the 'version'
field. During the encoding, it's the responsibility of the
API user to set a proper field of the union. I implemented
both f_PCUIF_ver_INFO_{Trxs,PDCHMask} helpers for that.
Version 9 is kept as default, so this change can be merged
independently of the actual implementation. We can bump
it and remove the compatibility glue once the new versions
of both osmo-bts and osmo-pcu are released.
Change-Id: Idf11bc4ba3ff0b00b32f2beab8fd020c67119d05
Related: SYS#4868, OS#4547
Immediate Assignment is sent on AGCH/PCH, while on PACCH we have
Packet Uplink/Downlink Assignment instead (see 11.2.29, 11.2.7).
Change-Id: I56473ae5dfa565d2c9d26984b14685a4163bf69a
When changing the PCUIF, we have to maintain backwards compatibility
with the release versions of both osmo-pcu and osmo-bts, not only
with the recent master. In order to achieve that, let's introduce
'mp_pcuif_version', so the PCUIF version can be adjusted.
Change-Id: I3cf7f908e606b91dd2cbddc168827dd074aed052
Related: SYS#4868, OS#4547
Fix confusion: f_ms_establish_ul_tbf() vs f_establish_tbf(). The
new name precisely describes what the function does. Sending a
RACH.ind and receiving a DATA.req with RR Immediate Assignment
is just a part of the TBF establishment process.
Change-Id: I1e22b9936bf68ebcaf70cefbc57345995ebbbaed
I noticed that TC_egprs_pkt_chan_req_reject_exhaustion has started
to fail since the refactoring change [1]. As it turned out, this
test case does not init a GprsMs instance, which is needed by a
function it depends on - f_TC_egprs_pkt_chan_req().
I don't really see the point of using the GprsMs abstraction for
such small test cases sending just a few messages, which do not
require us to keep any state. Let's just revert those changes,
and keep using the low level API.
[1] Ib3fee37580f0ea0530a659dec83656799bf57288
Change-Id: I4f32c138e7587e68d2d35d13d13c044893fec4b2
Let's test the code path where UL TBF is requested through DL ACK/NACK
here, since we already test the usual UL TBF through CCCH approach in
most tests.
rlc_mode is changed to ACKED since that's the mode we are using so far
in tests.
Change-Id: I5a9a2e8107c87fdbf74cc2f09ae5eeafbb13ad55