Closed 9600 closed 11 months ago
Are you sure the source-built versions match the packages? I built both package versions (LimeSuite tag v22.09.1, srsRAN release_22_04_LC01) from source, and I'm seeing the same behavior. This is without any of the custom flags debhelper
adds when building the package.
Building off of master (a470dfecd4963c6edb1952c2d32a713650a5f463 as of this comment), I saw this with both source-built and package-generated versions.
Let me re-rest this week and get back to you.
Were you able to verify the behavior? I rebuilt on a fresh Ubuntu 22.04 setup, and I'm still not seeing it with any of my boards.
With srsRAN now building on Ubuntu 22.04 I was able to test with this built from source, and then Lime Suite packages installed from the PPA (22.09.1), and I don't see these errors.
Not sure what the cause of the issue was when testing on Ubuntu 20.04 — and of course the srsRAN fork has now been updated to fix the build failure on Ubuntu 22.04 — but closing this issue.
Summary
The issue is that the libLimeSuite.so.22.09.1 which is built either via the MyriadRF PPA or locally with Debian packaging tools, results in _LMSSetSampleRate: Failed to set RX sampling rate and Too many trials being printed out.
Error details
andrew@lcran:~$ sudo srsenb /etc/srsran/enb.conf --- Software Radio Systems LTE eNodeB ---
Reading configuration file /etc/srsran/enb.conf... WARNING: cpu0 scaling governor is not set to performance mode. Realtime processing could be compromised. Consider setting it to performance mode before running the application.
Built in RelWithDebInfo mode using 22.04.0.
/build/srsran-tSvzof/srsran-22.04LC01/srsenb/src/enb_cfg_parser.cc:1507: Force DL EARFCN for cell PCI=1 to 1934 connect(): Connection refused Failed to initiate S1 connection. Attempting reconnection in 10 seconds Opening 1 channels in RF device=lime with args=index=0,rxant=LNAH,txant=BAND2,cal=all Supported RF device list: UHD lime file Number of requested channels: 1 Found device #0: LimeSDR-USB, media=USB 3.0, module=FX3, addr=1d50:6108, serial=00090726074D041C Reference clock 30.72 MHz Initializing limesdr device Setup RX stream 0 Setup TX stream 0 RX antenna/s set to: LNAH TX antenna/s set to: BAND2
==== eNodeB started === Type to view trace
LMS_SetSampleRate: Failed to set RX sampling rate
TX sampling rate: 3.84
Setting analog TX LPF BW to: 5.00
Setting manual TX/RX offset to 73 samples
Setting frequency: DL=1878.4 Mhz, UL=1783.4 MHz for cc_idx=0 nof_prb=15
Calibrating RX channel: 0, BW: 2.50
Calibrating TX channel: 0, BW: 2.50
Too many trials
Too many trials
^CStopping ..
Too many trials
Too many trials
Too many trials
Additional notes
Configuring cmake with _-DCMAKE_BUILDTYPE=RelWithDebInfo doesn't seem to trigger this problem, nor does stripping the library of symbols. Which were the only obvious differences I could see in how the library gets built as part of packaging.