The CI fails sometimes, because some processes/threads are created in
unpredictable order and if the client connects before the server is
ready some tests fail, as seen in the following log:
[UTIL] client/server iq-samples: launch sub-processes
[UTIL] threadCreate() for time client: creating thread (no affinity, default priority)
[UTIL] time manager configuration: [time source: server] [mode: client] [server IP: 127.0.0.1} [server port: 7374]
[UTIL] time client: connection to 127.0.0.1:7374 failed, try again in 1s
[UTIL] threadCreate() for time source iq samples: creating thread (no affinity, default priority)
[UTIL] threadCreate() for time server: creating thread (no affinity, default priority)
[UTIL] time manager configuration: [time source: iq_samples] [mode: server] [server IP: 127.0.0.1} [server port: 7374]
[UTIL] client/server: rlc ticks: 1000 pdcp ticks: 1000 x2ap ticks: 1000
CMDLINE: "/oai-ran/build/common/utils/time_manager/tests/test_time_manager_auto" "--time_management.mode" "server" "--time_management.time_source" "iq_samples"
[UTIL] client/server: rlc ticks: 0 pdcp ticks: 0 x2ap ticks: 0
CMDLINE: "/oai-ran/build/common/utils/time_manager/tests/test_time_manager_auto" "--time_management.mode" "client"
[UTIL] client-server iq-samples: ERROR
The client/server iq-samples test is started.
Then we see this log:
[UTIL] time manager configuration: [time source: server] [mode: client] [server IP: 127.0.0.1} [server port: 7374]
Which means that the client is running.
It tries to connect to the server (not running yet), but fails, so sleeps
for 1 second before attempting another connection to the server:
[UTIL] time client: connection to 127.0.0.1:7374 failed, try again in 1s
The sleep is done by the time_client thread, the main thread of the test
program continues and does a sleep(1).
Then the server arrives:
[UTIL] time manager configuration: [time source: iq_samples] [mode: server] [server IP: 127.0.0.1} [server port: 7374]
It sends all the ticks (to no client, since the client is still sleeping).
Then the test is over, the sleep(1) of the client finishes and the program
exits. The client never connected to the server.
We have this log:
[UTIL] client/server: rlc ticks: 0 pdcp ticks: 0 x2ap ticks: 0
And the test fails.
It's complicated to write an automatic test involving several uncoordinated
programs. The operating system may create processes in any order. We use
some sleeping to wait, but it may be too short or too long, leading to test
failure.
Let's see with more sleep if the CI fails less.
Fix PDSCH rate-matching for periodic CSI in UE
Changed the search structure to have periodic resoruce id in the outer
loop and configured resource set as inner loop to prevent configuring
resource that is not needed.
Use const for input pointers.
CI: refactor iperf3 UDP analysis function
- add helper function to covert reported bitrate values to Mbps for calculations
- clean up the UDP analysis function, improve test report formatting
- fix an issue where Gbps values were incorrectly reported
(example: https://jenkins-oai.eurecom.fr/job/RAN-SA-AERIAL-CN5G/4802/)
- update CI unit test report checks to match the new formatting
Rename the library to reflect that it's more a generic AAL driver than
specifically for AAL. To this end, rename some documentation.
As of this commit, the options are still called nrLDPC_coding_t2. The
next commit will correct this.
Co-authored-by: Xin Zhe Khooi <khooixz@comp.nus.edu.sg>
Changed the search structure to have periodic resoruce id in the outer
loop and configured resource set as inner loop to prevent configuring
resource that is not needed.
Use const for input pointers.
Add a CU-UP load tester and improve GTP performance
This adds a simple tester program for the CU-UP. It sets up a number of
UEs/bearers in the CU-UP via the E1 interface, then streams data in
downlink/uplink (as seen by the CU-UP). More information is available in
tests/nr-cuup/README.md.
Includes some fixes to properly verify that SCTP is installed, and use
"the right cmake way" to use it in cmake code (see commit for details).
simplify and make more local variables in nr_rx_pdcch() and fix small bugs...
simplify and make more local variables in nr_rx_pdcch() and fix small bugs
(size 9 RE instaed of 12 RE)
T hacks: add new tracer ant0 and a tool to plot ofdm data from usrp or rfsim
(output or input)
The tool ant0 dumps IQ data to a file (or stdout) as float complex or short.
It is needed for a student's project. (Well, we could use record/extract, but
it's simpler.) The tool ofdm-plot is used to plot output or input signal from
gnb in the frequency domain. May be useful to debug, why not.
numelt in place of config_isparamset
config_isparamset doesn't seem to work for array types, a temporary solution is
using numelt to verify if the parameter is set in config file
NR UE: fix reliable SIB19 reception when ntn-UlSyncValidityDuration-r17 expires
Fixes:
- gNB: fix scheduling of Rel-17 SIBs
- gNB: fix case in update_SIB1_NR_SI() when multiple Rel-17 SIBs would be defined
- NR UE: fix calculation of SI window expiry
Updating LEO scenario to test this
- set ntn-UlSyncValidityDuration-r17 to 5 seconds for LEO scenario
- NTN config: setting sib1_tda is not required anymore
Remove these F1-C port numbers. The configuration is not actually
applied, and the default F1-C port number 38472 (see F1AP_PORT_NUMBER)
is hardcoded in the F1 module.
For the moment, I leave the configuration option, in case we want to
allow to configure the port numbers later).
Make a common pattern, which makes it simpler to find them in gNB logs,
and, maybe more importantly, link tunnel creation and update (the update
message was missing the "incoming" TEID, so it is not easy to link to
creation apart from the UE ID, but even then the UE might have multiple
tunnels).
Reuse the gtpv1u_bearer_t for internal functions. Hopefully this will
also simplify implementation of IPv6, as it encapsulates IP address
handling within gtpv1u_bearer_t.
The (remote) port number of the GTP module is configured during
initialization. It is therefore not required when calling
newGtpuCreateTunnel(), and could even lead to bugs (if another port
number is used than specified during GTP init).
This is a simple CU-UP load tester. It sets up a configurable number of UEs in the
CU-UP, then sends GTP traffic in downlink/uplink in the N3-F1/F1-N3
direction at the CU-UP. It then compares the sent and received data.
Additional parameters, such as amount of traffic, test length, packet
size, and IP addresses/ports are configurable.
To run:
mkdir build && cd build && cmake .. -GNinja && ninja nr-cuup nr-cuup-load-test params_libconfig
./tests/nr-cuup/nr-cuup-load-test
./nr-cuup -O ../tests/nr-cuup/load-test.conf
The tester should be able to detect lost and reordered packets.
Example lost: Receive order 2,3,5,6 (lost packet 4)
=> on packet 2, expected 2, next=expected+1=3
=> on packet 3, expected 3, next=expected+1=4
=> on packet 5, expected 4, diff=5-4=1 (lost 1), next=payload+1=6
=> on packet 6, expected 6
=> 1 lost
Example reordered: Receiver order 2,5,3,6 (lost packet 4)
=> on packet 2, expected 2, next=expected+1=3
=> on packet 5, expected 3, diff=5-3=2 (lost 2), next=payload+1=6
=> on packet 3, expected 6, (reduce lost by 1 to 1), next=expected=6
=> on packet 6, expected 6
=> 1 lost
Example reordered: Receiver order 2,5,3,4,6 (no losses)
The boolean value v17_sib was not set to true when allocating si_schedulingInfo_v17 and schedulingInfo2_r17.
So multiple Rel-17 SIBs would lead to multiple allocations and only the last SIB being actually configured.
Fix this by removing the redundant v17_sib variable and directly check if the memory was already allocated.
Also move the definition and allocation of si_schedulingInfo_v17 closer to its actual usage.
Now that we reliably receive SIB19, we can set the ntn-UlSyncValidityDuration-r17 to a sane value
and thus make sure that the UE always has up-to-date SIB19 information.
A dedicated Msg3 TDA is simply not necessary: get_feasible_msg3_tda()
already iterates through all TDAs to find a suitable one for Msg3, and
takes things such as parameter Δ already into account.
The only prerequisite is that the ra_ResponseWindow is long enough. In
some configurations, Msg3 TDA ensured that Msg2 still happened during
ra_ResponseWindow, while Msg3 was outside of it. This is not necessarily
the case anymore, unless ra_ResponseWindow is long enough.
For the normal user, removing ra_ResponseWindow from the config will be
enough.
Since the parent commit, the ra_ResponseWindow will be chosen
automatically. Correspondingly, remove most ra_ResponseWindow options
from configuration files in the repository.
One problem in some mu=1 configs is that the ra_ResponseWindow=4 (10
slots), which might be overly restrictive if the L1 slot ahead is
somewhat close to 10 (e.g., 6), in which case we might not be always
able to allocate Msg2 in time.
A notable exception here is gnb-vnf.sa.band66.u0.25prb.nfapi.conf, in
which we still provide it (1) since it uses a non-standard configuration
of a ra_ResponseWindow longer than 10ms (see also issue #916 for more
information) and (2) as an example.
If the config file does not specify ra_ResponseWindow, select it to be
10 slots in mu=0, or 20 slots for all others.
This should give us enough flexibility to allocate Msg2 without running
out of the response window. The next commit will remove most
ra_ResponseWindow options in configs.
Scheduling for inactivity only deviates in details from the main
scheduling logic: we simply allocate a fixed number of RBs instead of
making it dependent of data.
In this commit, add an additional flag to sort UEs that are scheduled
for inactivity in front of "normal" UEs (with traffic). Then, schedule
those UEs by limiting/fixing the amount of RBs to schedule to
min_grant_prb.
Update the comparator() to properly sort UEs based on this flag and the
coefficient by applying a consistent logic of -1/0/1 (see man qsort) for
smaller/equal/larger. The inactivity flag, if set, will always have
higher priority over the coefficient to schedule these UEs first, as was
the case prior to this commit.
Remove sched_ctrl->sched_ulsch, and call the post-processing function
post_process_ulsch() in the places where sched_ctrl->sched_ulsch has
been assigned. This will have two effects:
- it is not necessary to have the final loop over all UEs, which might
be costly especially towards many UEs.
- till now, there is only a single sched_ulsch; however, a later commit
will allow to schedule multiple PUSCH transmissions from a single (DL)
slot. This would require to make sched_ctrl->sched_ulsch, and
also iterate through all sched_ulsch. In this new scheme, we can
simply call post_process_ulsch() multiple times, which is simpler (and
therefore, implicitly handleded as of this commit).
Write a single NR_sched_pusch_t structure in a single statement (where
possible). Assign to sched_ctrl->sched_pusch later, where the
post-processing functionality will be called in a follow-up commit.
Write a single NR_sched_pusch_t structure in a single statement (where
possible). Assign to sched_ctrl->sched_pusch in one place, where the
post-processing functionality will be called in a follow-up commit.
Avoid to overwrite the existing retInfo. Instead, create a new
NR_sched_pusch_t and assign to sched_ctrl->sched_pusch in one place. A
follow-up commit will directly trigger the post processor here.
Write a single NR_sched_pusch_t structure in a single statement (where
possible). Assign to sched_ctrl->sched_pusch in one place, where the
post-processing functionality will be called in a follow-up commit.
PF uses MCS and average throughput to decide which UE to schedule. Thus,
the first loop selects the MCS, and the second does the main UE
allocation after sorting by priority, depending on the MCS.
A follow-up commit will remove sched_pusch from NR_UE_sched_ctrl_t.
Thus, we cannot rely on sched_ctrl to store the MCS in the first loop,
and look it up from there in the second. Instead, store it as part of
the UE iterator data, and take it from there.
A later commit will remove sched_ctrl->sched_pusch. Instead, the TX
power is also stored for each HARQ process (as they contain
sched_pusch). Read it from there, which should have the same effect.
Robert Schmidt