It's not immediately obvious what numeric values mean for CPU scheduling policy.
Let's document this and add doc reference.
Related: OS#4107
Change-Id: Ib047762a336851e6205d77c83068a99d8a868e8b
Previously osmo-bts created /var/run/osmo-bts.pid which
isn't used by anything and makes it hard to run as non-root
user. Let's get rid of this pre-systemd relic.
Related: OS#4107
Change-Id: I86bcaedbc8cb1297476ad741eaa45585fea3c380
By default systemd will execute service with root directory
(or home directory for user instance) which might result in
attempts to create files in unexpected place. Let's set it
to 'osmocom' subdir of state directory
(/var/lib for system instance) instead.
Related: OS#4821
Change-Id: I4b7bcd441e5da81c4c5267715675041171a1ce1e
This helps understand the origin/purpose of those 2 bytes and
what's contained in them.
Related: SYS#5987
Change-Id: I607fdf6d627242e010fba35be1b9b0ffde820d08
This library will be used soon when adding Osmux support to osmo-bts.
Furthermore, it nice to have it available to make use of other general
interfaces to create connections, primitives, RTP and AMR related
functionalities, etc.
Related: SYS#5987
Change-Id: I49db4de715065c083e1249cbeae6298d6868e229
The PRIM_INFO_MODIFY is sent on receipt of the RSL MODE MODIFY message,
which may optionally contain new ciphering parameters. Before this
patch, osmo-bts-trx would ignore the new parameters and continue using
the old ciphering key/algorithm.
* Modify l1if_set_ciphering() to allow updating ciphering params.
* Call l1if_set_ciphering() when handling PRIM_INFO_MODIFY.
This problem was uncovered by BTS_Tests.TC_rsl_modify_encr, which has
been failing ever since it was introduced [1]. This patch makes it pass.
Change-Id: I2278703b487374c0de4cfc42b22e70aaf6548589
Related: [1] I4cbea499bb6a331d314e6573548a4540945208b5 osmo-ttcn3-hacks.git
Related: OS#3750
It will be used too by osmux code present in another file. This is a
preparation commit to simplify the one adding osmux support.
Change-Id: Ie7fa57bb04db9ad9b03971467e12ee7b8e4c190a
The function is long/complex enough, so having one extra struct in_addr
declared the function top only used in one specific small path to print
the variable is unnecesary.
Let's move it to the conditional path where it is used to print the
ip address.
Change-Id: I4c16bbca6a6779537517b6b196828b47eddaa516
There's no need to maintain a duplicate msgb_queue_flush(), which
returns the amount of freed messages (feature not used at all by the
callers).
Change-Id: I9841e18ca0b7b852130bbb02a510e43a3b3fd93f
The queue_limit_to method iterates the entire list of messages every
time a new message is added. Let's use msgb_{enqueue,dequeue}_count()
APIs to do that in constant time. It is true that since the queue is
limited to 1, there's usually at most 1 item in the queue so it's not a
real problem. However, when we add Osmux in the future, we may need to
tweak the amount of messages which can be in the list, due to Osmux
batching mechansim which may be more bursty sometimes.
In any case, this change doesn't make things worse for sure.
The patch also takes the chance to group the queue_limit_to + enqueue
into one function to avoid having the code spread several times.
Change-Id: I61818a3bb521c27bd21a8b6fa70581d27638ec9b
Make them more easy to read and edit by splitting to one element per
line when several elements are present.
Change-Id: I24ecfa1167b806dcb3a5a0c00343299df842a78b
The logic responsible for enabling and disabling the loop appears
to be broken: during my experiments trx_loop_amr_set() was never
called with loop=1. It's not clear what's the motivation behind
this breaker, so I propose to rip it out for now. This patch
finally makes the loop work (confirmed with a TEMS phone).
Change-Id: I09b649973d4269c4082a4fafa493c37825f95a9c
Related: SYS#5917, OS#4984
This reverts commit 1f9fbd8816, which
was a workaround rather than a proper fix. The root problem has
been fixed in libosmocoding [1], so ~50% BER is not the case for
AFS_SID_UPDATE frames anymore.
Related: I813081a4c0865958eee2496fe251ae17235ac842 (libosmocore.git)
Change-Id: Iebe139632daf1d1f72e39fe9d1497059f68625b9
We have two similar values in enum gsm0503_amr_dtx_frames:
* AFS_SID_UPDATE - precursor of SID UPDATE,
* AFS_SID_UPDATE_CN - the actual SID UPDATE.
The former is internally used by libosmocoding to mark the current
frame as a precursor of the actual SID UPDATE frame - the later.
+---+---+---+---+---+---+---+---+
| _ | _ | _ | _ | a | b | c | d | AFS_SID_UPDATE
+---+---+---+---+---+---+---+---+
| a | b | c | d | _ | _ | _ | _ | AFS_SID_UPDATE_CN
+---+---+---+---+---+---+---+---+
^
|
| We're here.
Do not treat the precursor frames as a SUB frames, as they contain
no information (rc=-1) and usually have *expected* BER ~50%. This
fixes unexpected RXQUAL-SUB > 0 in the Uplink measurements during DTX.
Change-Id: I46f32bbbcb6284615d05b8703945c5ca4da55b92
Related: SYS#5853
3GPP TS 45.008, section 8.3 defines active TDMA frame subsets for
TCH channels, which shall always be transmitted even during the
silence periods in DTX mode of operation. Each frame number
listed in this section corresponds to a single burst.
The Uplink measurements always contain TDMA FN of the *first* burst
of a block, so it does not make sense to match the given FN against
all FNs in the respective subset. Instead, we should match only
specific FNs in accordance with the block mapping rules defined in
3GPP TS 45.002, section 7, table 1.
In the active subset for TCH/F there is only one *complete* block
starting at FN=52. Incomplete blocks {52, 53, 54, 55} and {56, 57,
58, 59} contain only 50% of the useful bits (partial SID) and thus
~50% BER, so we don't treat them as SUB.
In the active subsets for TCH/H there are two *complete* blocks
for each sub-slot. Their respective first FNs can be efficiently
defined in a lookup table (see ts45008_dtx_tchh_fn_map[]). Note
that we can use a single lookup table for both sub-slots of TCH/H
because their TDMA FNs do not overlap.
This patch fixes unexpected SUB-RxQual values > 0 on TCH channels
with DTXu enabled and other than AMR (HR, FR, EFR) codec in use.
Change-Id: I8cc3a755a8ad4dc564439aab2298c1e97ac0592d
Related: SYS#5853
SACCH detection can be simplified by checking the RSL Link ID in
process_l1sap_meas_data(). This eliminates the need to lookup
the multiframe position by calling trx_sched_is_sacch_fn(), which
definitely takes more CPU time than just L1SAP_IS_LINK_SACCH().
Calling trx_sched_is_sacch_fn() is still required for BTS models
reporting the measurements via PRIM_MPH_INFO (legacy way),
separately from the related Uplink blocks.
This patch can be summarized as follows:
* detect SACCH and set .is_sub=1 in process_l1sap_meas_data();
** for PRIM_MPH_INFO use trx_sched_is_sacch_fn();
** for PRIM_PH_DATA use L1SAP_IS_LINK_SACCH();
* do not call trx_sched_is_sacch_fn() from ts45008_83_is_sub();
* modify the unit test - test_ts45008_83_is_sub_single();
** remove test_ts45008_83_is_sub_is_sacch().
Change-Id: I507e96ee34ac0f8b7a2a6f16a8c7f92bc467df60
Related: SYS#5853
* Print unexpected TDMA FNs, each on a separate new line.
* Do not abort() on mismatch, keep running.
Change-Id: I32c104b139f32a39c8347de19de6ad7caec811de
Related: SYS#5853
Attempting to detect missing Uplink frames in the *Downlink* path
is an ugly hack, that was needed before we had proper per-lchan
detection of missing bursts and NOPE indications in TRXDv1.
* [1] per-lchan detection of missing bursts was fixed in Aug 2018.
* [2] TRXDv1 support was introduced in Jul 2019.
Today is 2022 and it makes no sense to keep this hack. Even if
the transceiver does not support TRXDv1, per-lchan detection
will compensate the missing Uplink bursts.
Change-Id: Ic292d180ba64206fb4d88adb284f9f9d058b4587
Related: [1] I70d05b67a35ddcbdd1b6394dbd7198404a440e76
Related: [2] If61c71d20d590bf07bfd019afb33000a0b6135bd
The gitlog shows that these warnings were added in
744f745d7a by jolly (in 2012!).
As nobody has bothered for the past ten years, degrade them to comments.
Change-Id: Iebf9961cffbd7aa20b263f7dc0016a44782dec60
Related: osmo-bsc.git I9ef7e18f56aa86b48f0ffeec58406260736170f3
OsmoBSC assumed since If91d85331d402c3ab9c32b70c2c66cd7ba6ceb28 that
OsmoBTS has CCN (NACC, Network Assisted Cell Change). With OsmoBSC
Idf2d933aa8b03b1f708e56a08707fe6c620a97aa, instead of hardcoding some
features for OsmoBTS in OsmoBSC, only the features reported by the
BTS are used. Therefore report that BTS_FEATURE_CCN is supported in
OsmoBTS.
For previous OsmoBSC versions, this fixes the message:
Get Attributes Response: reported feature 'CCN' is not supported, while we thought it is.
Related: SYS#5922, OS#5538
Change-Id: Ic407e3be82afe61ed64b1cecf3dc94d2caff380a
I did not realize with my previous patch that there's a common place
where features available for all backends should be enabled. So move it
there.
Related: SYS#5922, OS#5538
Change-Id: Id688ddae5cd714355fcc5324f33c9397fecae261
According to osmo-bsc I0e80ca5afc06737273b6699bde6e325e454b57f6, the
PAGING_COORDINATION feature should be well supported in osmo-bts. Adjust
the bts_model_init functions to report this to the BSC, so it properly
sets this in SI13 and so we can avoid this notice in the log:
Get Attributes Response: reported feature 'PAGING_COORDINATION' is not supported, while we thought it is.
Related: SYS#5922, OS#5538
Change-Id: Ib06829501f4221a7f7b0344ec7991ceba74e7641
Currently (without this patch) if tch_dl_dequeue() yields nothing we're
scheduling dummy FACCH/H *regardless* if it's permitted to start at the
given TDMA FN or not. This may result in misaligned FACCH transmission,
so the MS will not able to decode anything and will report BER>0.
With this patch applied we schedule FACCH/H if it's allowed to start
at the current TDMA FN; otherwise send half-filled bursts with even
numbered bits contaning 232 encoded bits of the previous L2 frame
and 232 odd numbered bits all set to 0.
This patch does not guard against sporadic gaps in the Downlink TCH
queue. However when tch_dl_dequeue() constantly yields nothing, e.g.
when we end up with a codec mismatch, then this additional check saves
us from starting misaligned FACCH/H transmission. If we start at a wrong
TDMA offset, then all subsequent FACCH frames will be scheduled at wrong
offsets and thus none of them will be decoded by the MS.
Change-Id: I6f8af140a6ccf3d5fd7b98f6cb5c18e2c5e2f61b
Related: SYS#5919, OS#4823
* dl_ongoing_facch must be set for correct FACCH/H scheduling;
* dl_facch_bursts must be set for FACCH overpower to work.
Change-Id: Ief12eb67ad80de3b71f5226858dc2e0c8ae76948
Related: SYS#5919, OS#4823
If set, chan_state->dl_ongoing_facch indicates that we're sending
the 2 (out of 6) middle bursts of FACCH/H. In this case there is
no room for encoding an additional frame, because both even and
odd numbered bits are occupied by a previously encoded FACCH/H.
Right after calling tch_dl_dequeue():
* do not check the dequeued message against NULL,
* do not compare the message length to GSM_MACBLOCK_LEN,
* immediately free() it and jump to the sending part.
Change-Id: Ib923b8f5cc1063e9fc18acd2bdd003fd09f4b70f
Related: SYS#5919, OS#4823
Unlike SACCH, FACCH has no dedicated slots on the multiframe layout.
It's multiplexed together with TCH (speech or data) frames basically
by replacing (stealing) them. This is common for both TCH/F and
TCH/H, with the only difference that FACCH/H steals two TCH frames
(not just one) due to a longer interleaving period.
Let's implement the multiplexing in the common function, which is
used to dequeue to be transmitted frames - this slightly reduces
code duplication. Use a new name, so that it's clear what it does.
Change-Id: I9822b1a17185d5487f0f6d3ed0203e806c053d7d
Related: SYS#5919, OS#4823
Manual analysis of code didn't end up in finding any issue, so this
seems a false positive (I can really understand gcc failing to do proper
job here, this function has way to many jumps here and there.
"""
/sched_lchan_tchh.c:88:13: error: ‘rc’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
88 | int rc, amr = 0;
| ^~
"""
Change-Id: Ifebaee63a9dad04976ffb4438c32360687ef095a
Keep collecting C/I samples even if the MS has not yet applied
previously requested codec mode. The C/I ratio is expected to
stay more or less the same, no matter which codec mode is used.
Change-Id: Ieb5473ead7200f652b5d0e339e4e252d6567482d
Related: SYS#5917, OS#4984
With the current implementation the AMR loop is unable to upgrade
the current codec mode unless it reaches the worst possible value.
The problem is in the 'else' statement connection both 'if's.
Change-Id: I4c0fb28813373c3d4addd28c66f5136d2c4f9ed8
Related: SYS#5917, OS#4984
Max
Pau Espin
Vadim Yanitskiy
Harald Welte
Oliver Smith