Age | Commit message (Collapse) | Author | Files | Lines |
|
We have licensed the code under GNU Afffero Public License,
and state that in the first paragraph as well as in the link
to the license. However, a paragraph in the middle stated
"see the GNU General Public License", which is somewhat misleading.
Let's fix that.
Change-Id: I37e503b195fe43e1da42c080900504ca8e682e76
|
|
Change-Id: Iabc9b702e5f1513187e24f45d9ffe06ea940c3ec
|
|
Every BTS needs to have some graceful handling for the scenario
where it is time to send out a speech frame on TCH DL, but there is
no frame to be sent. One possible solution is to transmit dummy
FACCH, but this option is unattractive for TCH/AHS where FACCH
displaces two speech frames rather than one. A more elegant solution
is to emit a speech frame that is bad, causing the MS receiver to
declare a BFI condition to trigger substitution and muting procedure.
A bad frame is generated by gsm0503_tch_{afs,ahs}_encode() by setting
the payload length to 0.
Depends: libosmocore.git I82ce2adf995a4b42d1f378c5819f88d773b9104a
Related: OS#6049
Change-Id: I056f379715c91ad968f198e112d363a9009dc1c3
|
|
When adding support for Circuit Switched Data calls, we had to enlarge
the burst buffer size to accommodate bits for a maximum of 24 bursts.
Let's take advantage of this by utilizing the currently unused part of
the Uplink burst buffer for storing bits of previously decoded blocks.
This eliminates the need to allocate additional memory for SACCH.
Change-Id: I15047cd1df4476054b36f05616e41f5297d9bfe5
Related: SYS#5114, OS#4794, OS#4795, OS#4796
|
|
Change-Id: I7d89b2e50c7eeb54b734c4959eeeb1c63a51a315
|
|
Change-Id: I518a8ea268a9a6d48b04c291a03e5efbed5f571d
|
|
In accordance with 3GPP TS 44.021, sections 8.1.6 and 10.2.3, the
transmission of idle frames to the DTE is mandated when no data is
received from the radio interface. An idle frame has all data,
status, and E-bits to binary '1' (excluding the alignment pattern).
This requirement is currently implemented for the Uplink, see
function csd_v110_rtp_encode().
However, 3GPP TS 44.021 does not explicitly specify whether the same
rule is applicable to the Downlink, perhaps assuming a continuous
stream of bits on the CSD-over-TDM link. Currently, we transmit
a sequence of binary '0' on the Downlink instead of idle frames.
* In non-transparent (RLP) mode, whether all bits in a block are
set to binary '0' or '1' has no impact, as both scenarios lead
to an incorrect FCS.
* In transparent sync mode, any filling be it binary '0' or '1'
is perceived as incorrect or unexpected.
* In transparent async mode, it is more logical to transmit a
sequence of binary '1,' which will be interpreted as
a sequence of stop bits.
Let's align the Downlink with the Uplink for consistency and transmit
idle frames when no data is available for transmission. The modified
60-bit V.110 frames exclude the alignment pattern, so sending a
sequence of binary '1' is enough to achieve the intended goal.
Change-Id: I0b25cfac41b6d8dcf3bfd9d46d51a9665f1b047a
Related: OS#1572
|
|
Even though it might have a somewhat higher performance impact,
opting for the common code path for FACCH by allocating a msgb
on heap is more favorable for both readability and maintainability.
This choice is preferred over directly calling
gsm0503_tch_fr_encode() and then using a 'goto' statement.
A similar strategy will be adopted in an follow up patch for CSD.
Change-Id: I67cb5c6f4d15149996e17c78a59d66db396da8ff
Related: OS#1572
|
|
For the sake of consistency.
Change-Id: I16ce4c979c5b44fd67324eb2ed3da28a4b78221b
Related: OS#1572
|
|
Related: OS#4851
Change-Id: I3a2e3f94812cec0bbf7e3674172437fa359d014c
|
|
An MPH-INFO message is used to turn detection of uplink access bursts on
or off. This is required for voice group/broadcast channels.
Related: OS#4851
Depends: libosmocore.git Ibd6a1d468a70126a8f67e944fcb916969cc3c36b
Change-Id: I9045437d52984b7abe00fbc815d7f83c62c0fb5a
|
|
- If the llist is flushed during rx rsp callback, when the flow is
returned to trx_ctrl_read_cb() it would access tcm which was in the
llist and end up in use-after-free.
- We need to store state on whether code path is inside the read_cb in
order to:
-- Delay transmission of new message if callback calls trx_if_flush()
followed by trx_ctrl_send(), since the trx_ctrl_send() at the end of
trx_ctrl_read_cb would retransmit it again immediatelly.
-- Avoid accessing tcm pointer if the callback called trx_if_flush(),
since it has been freed.
Related: OS#6020
Change-Id: Ibdffa4644aa3a7d219452644d3e74b411734f1df
|
|
This reverts commit 4444262a6ab1e1e231ea81c4ec990f1a1f571a1f.
This commit introduced several side effects:
- tcm is left out of the l1h->trx_ctrl_list and hence won't be ever
retransmitted.
- Since tcm is removed before rsp callback, the llist may become empty
and if somehwere in the rsp callback a new message is enqueud it will
be sent immediatelly, and will be retransmitted again when
trx_ctrl_read_cb() calls trx_ctrl_send() at the end.
Change-Id: Ideb2d08ac8a2902bceeabfb055c59c9a13dbe3c0
Related: OS#6020
|
|
Change-Id: Idba18c217187f7aacc2fcb0c330814f3802a9306
|
|
Change-Id: I88c6c6af18be054bd152832e60c8afbbd16098a3
Depends: libosmocore.git Ia4208e10d61843dd6ae77398f6624c918dc81ea4
Depends: libosmocore.git I85316af9b57e8113077305798cb2d82a24e48e00
|
|
The RSL link is configured/set up by the BBTRANSC NM object, hence move
it to the appropiate substruct.
Related: OS#5253
Change-Id: I62937cbd81c27274b9f5f70d454d5319a6898c7b
|
|
Other bts models like sysmo,lc15,oc2g properly handled
rel_act_kind=LCHAN_REL_ACT_REACT under the bts_model_lchan_deactivate()
implementation, but bts-trx didn't.
As a result, when BCCH_INFO(SYSINFO_TYPE_3) coming from BSC (RSL)
containing a different BS_AG_BLKS_RES triggers CCCH re-activation,
it would only deactivate it but not re-activate it.
That means no SIs were being scheduled if bts was configured with
"channel-descrption bs-ag-blks-res 2" in osmo-bsc.cfg, and phones would
not see the cell.
Related: OS#1575
Change-Id: I61e1681fbaa2c993b529d58b581c99166b62bda3
|
|
UPLINK ACCESS (RACH on TCH) was enabled for osmo-bts-sysmo only. Now it
is also enabled for:
* osmo-bts-lc15
* osmo-bts-oc2g
* osmo-bts-trx
* osmo-bts-virtual
Change-Id: Iae0db6bfcf6629c114436a79648e832c82835abe
|
|
Change-Id: I95d4e4ee3938cfabc1695959cc82a1efbbf0d7ed
Related: OS#1572
|
|
It may happen that only FACCH is available for transmission, so msg_tch
would be NULL in this case. Check it before dereferencing.
Change-Id: I0e7d5634b5223bc246badbb8e94b620c967ab121
Related: OS#1572
|
|
In f2c902c2 I accidentally bumped the logging level for PDCH decoding
errors to LOGL_NOTICE, making osmo-bts-trx spam the logging with
hundreds of 'Received bad data' messsages. Revert this.
Change-Id: Idb963f1a779dfa172825f6d481740cb0c4165485
Fixes: f2c902c2 "osmo-bts-trx: unify and enrich 'Received bad data' logging"
|
|
Change-Id: I7bb341867e362bf2061608ff54c3596ad209af90
|
|
... for the sake of consistency with rx_tchh_fn().
Change-Id: Ie331da78eb3831e35d255583466e0d09b093b132
|
|
Change-Id: I7902f382e1d83ef9ad2cf6f92e31eeb16f6b797c
|
|
Change-Id: I373dbbc3d427858f76d07ff85633e07fe2600770
Related: OS#1572
|
|
Change-Id: I31c59a78a362986b0cd1bc6650c325ee9f206519
Depends: libosmocore.git I4685376c8deb04db670684c9ebf685ad6fc989fa
Related: OS#1572
|
|
The FACCH/[FH] coding rules are specified in 3GPP TS 45.003, sections
4.2 and 4.3. The key difference is that unlike with speech channels,
FACCH does not replace data frames completely, but disturb a fixed
amount of bits fom them. This is why we need to use separate
gsm0503_tch_[fh]r_facch_{en,de}code() API for data channels.
Change-Id: I4c6736e84c271240d457998de688c0baf59fe578
Depends: libosmocore.git I0c7a9c180dcafe64e6aebe53518d3d11e1f29886
Related: OS#1572
|
|
* enlarge the maximum burst buffer size to 24 * (2 * 58) bytes;
* enlarge per-l1cs Uplink burst mask to hold up to 32 bits;
* enlarge per-l1cs Uplink meas ring buffer to 24 entries;
* add new meas modes: SCHED_MEAS_AVG_M_{S22N22,S24N22};
Change-Id: I08ffbf8e79ce76a586d61f5463890c6e72a6d9b9
Depends: libosmocore.git Ib482817b5f6a4e3c7299f6e0b3841143b60fc93d
Related: OS#1572
|
|
The goal is to unify encoding functions in tx_tch[fh]_fn(), so that
in a follow-up change adding CSD we could use a switch statement.
Change-Id: I15318e497b236128f779769e4fa99f307ea431ea
Related: OS#1572
|
|
This commit aims to make the argument naming/ordering more consistent:
* rename l2/l2_len and tch/tch_len to data/data_len;
* common arguments first (consistent ordering), specific last.
Change-Id: I3d8c90b82f2a55b0c5c2d6b4efb8fd962508534d
Related: OS#1572
|
|
With this change the application of ECU in the uplink path becomes
consistent across all OsmoBTS models, enabled or disabled per vty
config setting "rtp internal-uplink-ecu". An additional behavioral
change from the previous trx-model-only implementation is that ECU
insertion is now done after the link quality check in l1sap, thereby
fixing the bug where this quality check would sometimes suppress
ECU output and replace it with BFI markers in RTP.
In the new implementation when the internal ECU is enabled and
available for the selected codec (currently FRv1 only), the RTP output
will gap (standard representation of BFI in RTP) only during DTXu
pauses as indicated by a received SID frame (either valid or invalid),
and the SID frame that triggers the switch from ECU mode into pause
mode is reliably emitted in RTP.
Related: OS#6040
Depends: I3857be84bba12aaca0c2cca91458b7e13c5a642a (libosmocore)
Change-Id: Iac577975c9ab50cb8ebbc035c661c1880e7cecec
|
|
In preparation for moving the now-optional application of ECU in UL
path from osmo-bts-trx model-specific code to the common layer,
move ECU state allocation and freeing from trx model to l1sap.
Related: OS#6040
Change-Id: Ic98a2eb26b5a99bc4a89ad07ae87c9a86b921418
|
|
Current osmo-bts-trx includes a provision for invoking ECUs from
libosmocodec in the UL path from the channel decoder to the RTP
output. This pre-existing implementation is counter to the spirit
of 3GPP specs (a BTS should merely mark BFI conditions in its UL
output, as opposed to actively modifying the frame stream with an ECU),
inconsistent between different osmo-bts models (only in -trx and no
others), and inconsistent even within osmo-bts-trx itself, where
the link quality check in l1sap will sometimes suppress the output
of the ECU - a quirk which the designers of the current mechanism
most certainly did not intend.
The solution decided upon in OsmoDevCall on 2023-06-21 is to make
this ECU optional per vty config, and move it from the trx model
to the common layer to resolve the inconsistencies. Implement the
first part: make the ECU application optional per vty config.
For backward compatibility with existing deployments, the new
"rtp internal-uplink-ecu" setting is enabled by default on osmo-bts-trx
but not on any other models.
Related: OS#6040
Change-Id: I0acca9c6d7da966a623287563e0789db9e0fae8e
|
|
In all Uplink lchan handlers we do memset() the Rx burst buffer on
bid=0 and there is no need to do that again for NOPE.ind.
Change-Id: Ia6803b8606d86bd2d011fe21f7a540d2115aa654
|
|
In my recent patch a0770250, among with the new burst buffer
allocation/release strategy, I introduced a regression:
/* send burst, if we already got a frame */
- if (br->bid > 0) {
- if (!*bursts_p)
- return -ENODEV;
+ if (br->bid > 0)
goto send_burst;
- }
We used to allocate the burst buffers in Rx/Tx lchan handlers, and
release them in case of an error, e.g. when no block is available
for transmission. In the case of Tx burst buffers, the state of
Tx burst buffer was additionally used to check if we have a valid
Tx block for transmission, as can be seen in the code snippet above.
As a side effect of my patch, osmo-bts-trx now keeps transmitting
3 out of 4 bursts (br->bid > 0) of the last valid block, until the
next valid Tx block is available for transmission.
This problem was not affecting the CS domain, where it's expected to
have a more or less constant pressure of Tx blocks. However it did
show up in the PS domain, where in the absence of active TBFs the
PCU may omit DL blocks.
Add a new field 'dl_mask' to struct l1sched_chan_state, similar to
the existing 'ul_mask', and use it to reconstruct the removed logic.
Change-Id: I4538a8fe6b29f8d6eca33ad27d4a9852e3a3e86c
Fixes: a0770250 "osmo-bts-trx: alloc/free burst buffers in trx_sched_set_lchan()"
|
|
Change-Id: If3c2d9d93399411d4466779ce8e8987843b521b1
|
|
In speech TCH operation, there will always be times when a speech frame
needs to be transmitted on the downlink, but there is no frame available
to transmit (gap in the incoming RTP stream), or the logic of DTXd says
that no frame shall be transmitted at this FN, but we are not doing
physical DTXd. Previous osmo-bts-trx code sent dummy FACCH during such
conditions, but this approach is bad for TCH/HS where we would like to
transmit good speech frames or speech frame gaps one 20 ms frame at a
time, rather than take out pairs of frames for dummy FACCH/H. Other
BTS models (sysmoBTS PHY) send out invalid speech blocks with inverted
CRC3 under the conditions in question - do the same in osmo-bts-trx.
The present change modifies osmo-bts-trx TCH DL frame gap behavior
only for TCH/FS, TCH/HS and TCH/EFS; for all other channel modes,
including AMR speech, the previous behavior of sending dummy FACCH
is left unchanged.
Depends: Iade3310e16b906efb6892d28f474a0d15204e861 (libosmocore)
Change-Id: I78106802a0aa4af39859c75d29fe0e77037899fe
|
|
Change-Id: I382b9994db01d58515c89c8de6250cd3239b8861
Related: OS#1572
|
|
There is no more need to access the UL/DL buffers via pointer to pointer
thanks to the previous patch [1] moving the memory management routines
out of the logical channel specific Rx/Tx handlers.
Change-Id: Ib1e8e86ac60a7ac9b0415ef66b609eaa33443c19
Related: [1] c45e03726af8a2b46b7166b3c47bc666ba933a69
|
|
It's a lot cleaner if the Rx/Tx burst buffers are allocated and free()d
in one place rather than in each lchan handler individually. Allocate
the buffer(s) during the lchan activation, free() on deactivation.
Regardless of lchan type, allocate the maximim size of 4 * (3 * 2 * 58)
bytes, which is sufficient to store 4 8PSK modulated bursts. This way
we reduce the load on the memory allocator at the cost of memory
consumption, what should not be a big problem.
Change-Id: I6a5f76023fc492786076a63016f81285b3576c33
Depends: libosmocore.git I1c38ccc2b64ba9046bb3b1221d99cc55ec493802
Related: OS#1572
|
|
The new vty option "rtp hr-format (rfc5993|ts101318)" selects the
RTP output format to be used for HR1 codec, consistently across
all models. The default is set to match legacy behavior: ts101318
on osmo-bts-{lc15,oc2g,sysmo} and rfc5993 on osmo-bts-trx.
On models where no legacy behavior is applicable, the default is
set to rfc5993 as a forward-looking measure - see OS#6036.
Closes: OS#5688
Change-Id: I168191874a062429a57904511a1e89e3f588732e
|
|
As a preliminary step before making the RTP output format from
OsmoBTS configurable with a vty option, we need to make the internal
format consistent across all models. The "natural" output from a
"pure" GSM 05.03 channel decoder is TS 101 318, which is also the
output format of all currently supported proprietary PHYs - adopt
it as our internal format.
Related: OS#5688
Depends: I6e75ca95409b5c368e8e04d0e0aba41e0331d9e6 (libosmocore)
Change-Id: I41bce6226964975cb85aea89e4c0f9e11e4929b8
|
|
As of commit 1160cabefb23, the common layer of osmo-bts accepts both
TS 101 318 and RFC 5993 formats for HR1 codec, and always passes
the more basic TS 101 318 format to the BTS model. Unfortunately,
osmo-bts-trx has model-specific code checking the payload length
that was overlooked in that patch, causing breakage. Fix that bug.
(The actual channel encoding function in libosmocoding already
accepts 14-byte payloads.)
Related: OS#5688
Fixes: I702e26c3ad5b9d8347e73c6cd23efa38a3a3407e
Change-Id: I0e251faeffb76d2604a4100c848141d239d1d86f
|
|
As detailed in OS#6033, osmo-bts-trx was emitting its own invented
packet formats to signal BFIs (bad frame indications), different
from the vty-configured ("rtp continuous-streaming" or not) BFI
signaling method implemented in l1sap and used by all other BTS models.
In the case of EFR codec, the made-up BFI format previously used by
osmo-bts-trx caused EFR operation to be broken: a spec-compliant EFR
decoder on the receiving end of the RTP stream (a network transcoder
or the MS on the other GSM call leg) would receive and decode garbage
EFR frames of 244 zero bits instead of invoking the spec-defined bad
frame handler, causing bad audio artifacts for the user. The same
situation would also happen for FR1 codec, but the application of
ECU masked this bug: with the ECU invoked in the UL output path,
the BFI emitting code for FR1 never executed.
In the case of AMR and HR1 codecs, the model-specific BFI packet
format of osmo-bts-trx is currently harmless, but:
* The extra code adds unnecessary complexity;
* BFI packet formats are inconsistent between osmo-bts-trx and
other OsmoBTS models;
* BFI format is even inconsistent within osmo-bts-trx itself, as
under certain conditions the common code in l1sap will override
the UL payload from the BTS model and emit its own form of BFI
instead.
Fix all of the above by removing trx model-specific BFI formats
for all codecs, and let l1sap handle all BFIs.
Related: OS#6033
Change-Id: I8f9fb5b8c5b2cad4b92ac693c0040779f811981a
|
|
Change-Id: Ib0bed7a4196305e2b413ca657f4fa8fb6f0d39f6
|
|
In all OsmoBTS models in FR and EFR speech modes, as the UL traffic
frame stream is passed from the PHY to l1sap and ultimately to RTP
output, all passed frames are checked for SID. The Boolean result
of this SID check is used in two ways:
* RFC 3551 instructs that the Marker bit in the RTP header SHOULD be
set in the first packet of a talkspurt after a period of silence
suppression. OsmoBTS implements this recommendation by setting
the Marker bit in the first RTP-transmitted non-SID UL packet
that was preceded by SID.
* In osmo-bts-trx only, this same SID check serves a second purpose:
handling of BFI conditions depends on whether or not the last received
good traffic frame was a SID. If a BFI condition is received after
a non-SID traffic frame, the ECU is invoked, and if this ECU fails
or when libosmocodec does not provide an ECU for the codec in use,
a special FR (EFR) BFI frame of 260 (244) zero bits is emitted.
Both the ECU call and the peculiar form of BFI are skipped when the
last received good traffic frame was a SID.
For both of the above purposes, in all BTS models, change the SID check
from osmo_{fr,efr}_check_sid() to osmo_{fr,efr}_is_any_sid(). The
effect of this change is that "what counts as a SID" for the just-listed
purposes changes from recognizing only perfectly uncorrupted SID frames
without any bit errors to recognizing all frames which the rules of
GSM 06.31 (FR) or 06.81 (EFR) classify as either valid or invalid SID.
Change-Id: I5c24b379deda8ae551f9924d10770da50e6acbbd
|
|
osmo-bts-trx uses the RFC5993 RTP payload format, so when handling the
RTP payload use GSM_HR_BYTES_RTP_RFC5993, instead of GSM_HR_BYTES + 1
Depends: libosmocore.git I125ef9cdab98c073971841c175b1a7dcd927f9c2
Related: OS#5688
Change-Id: I54dd3adab88e2262913f7b1e89340a0246c88a8a
|
|
- Add out-parameter to enable returning a NACK cause (ignored if NULL)
- Return appropriate NACK cause if TRX number is unknown (fixes OS#5967
together with change I37e6b23ed95260a8188910cf9754faffcba519c5)
Change-Id: If734ea2c8cae4c1f99b02520dffa4e3862a67745
Related: OS#5961, OS#5967
|
|
Change-Id: I684482064136a461d01cace3cd37afc8b68458cc
Related: OS#5994
|
|
Change-Id: I5fd1d17da09a5f0eee3d69fcb4788c106a240e21
Related: OS#5994
|