see https://gitlab.eurecom.fr/oai/openairinterface5g/issues/227
When the UE connects to the eNodeB and receives its IP address from the
network, it calls system() to set it in the linux kernel world. This call
is not done in a realtime thread, but in the NAS, which uses its own thread,
independent of the realtime processing.
In some situations this totally disrupts realtime processing.
It is difficult to know precisely why that happens, but it seems that calling
fork(), as system() does, in a multi-threaded program is not a good idea. (So
say several people on the internet.) It is not clear why the softmodem is
impacted, but it seems that fork() is really what triggers the disruption.
Several tests lead to that conclusion.
To fix the problem, we create a child background process very early in main()
(before anything else basically). Then instead of calling system(), the main
process sends the string to the background process. The background process
gets the string, passes it to system() and reports the success/failure back
to the main process.
This solution involves a lot of system calls, but calling system() in the
first place is not cheap either. As long as no realtime thread uses this
mechanism, things should be fine. Time will tell.
Develop integration w11
Summary of changes:
- bug fixes
- RCC/RRU in automatic test setup
- more work on UE/MIMO
- LimeSDR support (5/10MHz)
- compilation under ubuntu 16.04, in particular kernel modules (nasmesh, ue_ip)
See merge request !148
the argument 'coded_bits_per_codeword' has to be an array in
case of several codewords.
The calling sites have been adapted.
Today, only the first index is used, so calling sites where
'coded_bits_per_codeword' is an integer pass the address
of it. It is expected that 'dump_dlsch2' will check in the
future that there is one or two codewords and only access
'coded_bits_per_codeword[1]' when it's sure there are
really two codewords.
On some hosts, compilations with the T tracer was failing.
The error was:
common/utils/T/T.h:15:19: fatal error: T_IDs.h: No such file or directory
The problem was that in CMakeLists.txt some targets depend on the
pre-generation of T_IDs.h but this dependancy relation was not set,
so those targets could be generated before the T (it was the case
with HASHTABLE).
This commit fixes that. Basically, we take all the targets found in
"add_executable" and "add_library" and make them depend on the
T if the T is enabled. Almost all existing targets were added,
even those that may not need it.
The problem of this approach is that someone adding a new target
using the T will not necessarily add a dependancy there.
Another solution would be to generate T_IDs.h at "cmake" stage,
not "make" stage and use an "if (NOT EXISTS T_IDs.h)" to generate
the file. We lose the dependancy relation though.
Things may be changed if maintenance cost is too high.
768*8 is too small, in 20MHz, the phy-test mode
(in default mode?) has a buffer of size 6378.
Let's set the buffer size to the input's one plus 4 bytes
(to accomodate for CRC, not sure about this).
New tests: 0186xx for TDD with the huawei 3276 dongle.
All necessary files to configure and use it have been added.
Existing code has been adapted for the TDD tests.
Only the test setup v2 has been adapted.
The file test_case_list.xml has been changed to add tests
0186xx, but only to work with test setup v2. It won't
work with test setup v1. To be completed if needed.
The tests are similar to FDD tests.
Throughputs for TDD have to be adjusted, the throughputs
of FDD are used for the moment.
Cedric Roux