When a container image already has cached the first apt-get update, subsequent calls to apt-get install might fail in case repo information changes as is the case in https://github.com/CloudNativeDataPlane/cndp/blob/main/containerization/docker/ubuntu/Dockerfile#L48. The Dockerfile needs to be updated so the required packages are installed wiht apt-get update, or just move them to the first apt-get update && apt-get install line.

Difficulty: Fairly hard as one needs to learn CNDP graph node library and GTP-u protocol.

Description:
GTP-u graph node for CNET a CNDP graph node-based network stack (IPv4/UDP/TCP)

• Create and test a GTP-u graph node for the CNET stack. CNET stack has a stub version for a starting point cndp/lib/cnet/gtpu.
• Graph Node library source and documentation.
  • The graph node library was taken from DPDK with minor changes to work within CNDP.
• GTP-u graph node needs to comply to GTP-u protocol.
• Needs to be tested and verified it works with the CNET stack.
• This one can be fairly complex as one needs to learn the graph node library and the GTP-u protocol.

TODO: Add more information.

Investigate if Fedora and Ubuntu place the CNDP libraries in the same/expected locations

Update .clang-format file to include the new option InsertBraces: false when clang 15 is available in Fedora + Ubuntu.

https://clang.llvm.org/docs/ClangFormatStyleOptions.html

this will add the check for {} in single line statements

Difficulty: Hard

Add IPv4 fragmentation/reassembly to the CNET stack. This is not a new graph node, but a set of function handle the RFC-791.

Enable a user to load their own bpf program on the interface being used for AF_XDP.

Note: this is partially enabled through the bpf map pinning support. Perhaps Also document this better to allow the end user to use whatever their preferred BPF loader is.

Hello,

I'm analyzing porting dpdk apps to cndp. We need jumbo frames (up to 9KB) and we use chaining pktmbuf functions.
It seems that XDP supports jumbo frames using multi-buffer. Is it planned to support it in CNDP ?
Moreover, when do pktmbuf chaining capability would be available please ?

Thanks in advance for your answer.

Niry

Had issues with quicly/picotls detection and even if I manage to add workarounds, the build will stop with error that signals quicly API have changed, I suspect that version of quicly is too new and example code needs to be updated.

Hello, I am trying the cndpfwd application in fwd mode,
I'm generating packet using T-Rex Traffic Generator with the src/dst mac address are:
- dest_mac: 40:a6:b7:19:f2:39 src_mac: 40:a6:b7:19:f2:38
My current fwd.jsonc configuration for the 2 ports of the cndpfwd app is:
"lports": {
"enp175s0f0:0": {
"pmd": "net_af_xdp",
"qid": 0,
"umem": "umem0",
"region": 0,
"description": "LAN 0 port"
},
"enp175s0f1:0": {
"pmd": "net_af_xdp",
"qid": 0,
"umem": "umem0",
"region": 1,
"description": "LAN 1 port"
}
Traffic was generated to port enp175s0f0. The app forwarded the traffic back to this port as it cannot find the corresponding lport as said in the document
"In forward mode (‘fwd’), the destination logical port on which to forward the packet is specified by the last octet of the destination MAC address. It should be a valid lport number 0-N. If the destination port does not exist the packet is sent back out the port on which it was received"

I would like to know how to configure the second lport number of (enp175s0f1) so that the cndpfwd app can forward the packet to this one according to the last octet 39 of the destination MAC address?

Thanks,

Difficulty: Medium

The Message Channel (MsgChan) is a library utilizing lockless rings to send/receive 8 byte messages between threads in a single process. The library creates two lockless rings one for sending and one for receiving messages. The current usage of Message Channels is to send messages between Go language threads and the CNDP C code threads. The library is creating and maintaining these lockless rings plus given the developer a simple set of APIs to utilize the channel.

The goal of this task is to modify the library to allow sharing the lockless rings between processes. The messages can not be pointers to other shared memory regions only offsets into the shared memory. The messages can be any 8 byte message and does not need to point to some other memory. The lockless rings appear to already be shared between processes, but the MsgChan structures are not completely sharable.

The implementation should rely on Linux IPC and existing shared memory infrastructure wherever possible.

Have any evaluations been done on using lwIP as an alternative to the IP stack that was created from scratch?

If so, why was this design chosen instead? Was it due to performance reasons?

Hello there!

I am interested in finding out more information about the timeline for IPv6 support. If this becomes available, we would be eager to test it out for these use cases:

1. lw4o6 (RFC 7596) including both:

• AFTR
• B4

2. Keyed IPv6 tunnel (RFC 8159)

We would be happy to contribute this code back as the example projects.

Any feedback would be greatly appreciated.

Excuse me, can I use WireGuard CNDP as the client to connect to the wireguard server supported by the normal Linux kernel? 

Example:
Wireguard cndp<———->normal WireGuard server.

I am currently testing in this way and cannot receive any data packets on the server.

It appears while adding txgen latency support, the packet data offset with pktmbuf_mtod() is not pointing to the start of packet data as expected. It appears libbpf 0.6.0 behaves correctly and we need to move to libxdp to verify the problem exists.

Difficulty: Easy

Enhance the test-cne infrastructure to be more consistent or easier to use. Need all of the tests to look similar and produce similar output.

For example to show this test is different then the others. This test uses printf and not the tst_info() or tst_error() type APIs, these need to be cleaned up to use the correct APIs from the tst_ XXX library.

Difficulty: Hard

Create a new CNDP binding layer to allow Python scripts to utilize CNDP libraries. The new binding layer needs to be similar to the Go/Rust binding layers. Create the bindings in lang/python. As an example to use the bindings, create a Python application that uses the bindings similar to the Go fwd example.

Preferably the bindings are auto-generated as much as possible.

The privileged flag was removed for lports in the c CNDP libraries in this PR and a new flag called xsk_pin_path was also added.

The privileged flag is automatically configured if the xsk_pin_path or the xsk_uds configurations are configured,. the same should be done in rust

Difficulty: Medium to easy may have to learn Go language

Create a new example Go application doing some real world task and use the CNDP Go Binding layer. We currently have a simple loopback/forward like application called fwd along with a set of Go test code to verify Go bindings work.

Pick a reasonable Go application doing some type of work other than simple forwarding. Write the Go code and utilize the CNDP Go binding APIs. May need to add new or update Go binding APIs.

Split from #232 since there're multiple tasks.

The Go Binding API testing covers 45.8% of statements, needs to add more unit tests to increase code coverage.

Using Ubuntu 20.04 setup using ansible

"/home/canopus/cndp/tools/cne-build.sh" clean static build
Build environment variables and paths:
CNE_SDK_DIR : /home/canopus/cndp
CNE_TARGET_DIR : usr/local
CNE_BUILD_DIR : /home/canopus/cndp/builddir
CNE_DEST_DIR : /home/canopus/cndp
PKG_CONFIG_PATH : /usr/lib64/pkgconfig
build_path : /home/canopus/cndp/builddir
target_path : /home/canopus/cndp/usr/local

*** Removing '/home/canopus/cndp/builddir' directory

Static build in '/home/canopus/cndp/builddir'
Release build in '/home/canopus/cndp/builddir'
Ninja build in '/home/canopus/cndp/builddir' buildtype='release'
The Meson build system
Version: 0.53.2
Source dir: /home/canopus/cndp
Build dir: /home/canopus/cndp/builddir
Build type: native build
Program cat found: YES (/bin/cat)
Project name: CNDP
Project version: 22.08.0
C++ compiler for the host machine: c++ (gcc 9.4.0 "c++ (Ubuntu 9.4.0-1ubuntu120.04.1) 9.4.0")
C++ linker for the host machine: c++ ld.bfd 2.34
C compiler for the host machine: cc (gcc 9.4.0 "cc (Ubuntu 9.4.0-1ubuntu120.04.1) 9.4.0")
C linker for the host machine: cc ld.bfd 2.34
Host machine cpu family: x86_64
Host machine cpu: x86_64
Checking for size of "void *" : 8
meson.build:91: WARNING: Consider using the built-in debug option instead of using "-g".
Has header "execinfo.h" : YES
Found pkg-config: /usr/bin/pkg-config (0.29.1)
Did not find CMake 'cmake'
Found CMake: NO
Run-time dependency libexecinfo found: NO (tried pkgconfig and cmake)
Run-time dependency numa found: NO (tried pkgconfig and cmake)
Run-time dependency libbsd found: YES 0.10.0
Run-time dependency json-c found: YES 0.13.1
Run-time dependency libpcap found: YES 1.9.1
Run-time dependency libnl-3.0 found: YES 3.4.0
Run-time dependency libnl-cli-3.0 found: YES 3.4.0
Run-time dependency libnl-route-3.0 found: YES 3.4.0
Run-time dependency libxdp found: NO (tried pkgconfig)
Run-time dependency libbpf found: YES 0.3.0
Has header "linux/if_xdp.h" : YES
Has header "bpf/xsk.h" : YES
Has header "bpf/bpf.h" : YES
Run-time dependency libbpf found: YES 0.3.0
Compiler for C supports arguments -muintr: NO
Run-time dependency libdlb found: NO (tried pkgconfig)
Compiler for C supports arguments -Wno-pedantic -Wpedantic: YES
Compiler for C++ supports arguments -Wno-pedantic -Wpedantic: YES
Compiler for C supports arguments -Wcast-qual: YES
Compiler for C++ supports arguments -Wcast-qual: YES
Compiler for C supports arguments -Wdeprecated: YES
Compiler for C++ supports arguments -Wdeprecated: YES
Compiler for C supports arguments -Wformat-nonliteral: YES
Compiler for C++ supports arguments -Wformat-nonliteral: YES
Compiler for C supports arguments -Wformat-security: YES
Compiler for C++ supports arguments -Wformat-security: YES
Compiler for C supports arguments -Wmissing-declarations: YES
Compiler for C++ supports arguments -Wmissing-declarations: YES
Compiler for C supports arguments -Wmissing-prototypes: YES
Compiler for C++ supports arguments -Wmissing-prototypes: NO
Compiler for C supports arguments -Wnested-externs: YES
Compiler for C++ supports arguments -Wnested-externs: NO
Compiler for C supports arguments -Wold-style-definition: YES
Compiler for C++ supports arguments -Wold-style-definition: NO
Compiler for C supports arguments -Wpointer-arith: YES
Compiler for C++ supports arguments -Wpointer-arith: YES
Compiler for C supports arguments -Wsign-compare: YES
Compiler for C++ supports arguments -Wsign-compare: YES
Compiler for C supports arguments -Wstrict-prototypes: YES
Compiler for C++ supports arguments -Wstrict-prototypes: NO
Compiler for C supports arguments -Wundef: YES
Compiler for C++ supports arguments -Wundef: YES
Compiler for C supports arguments -Wwrite-strings: YES
Compiler for C++ supports arguments -Wwrite-strings: YES
Compiler for C supports arguments -Wno-address-of-packed-member -Waddress-of-packed-member: YES
Compiler for C++ supports arguments -Wno-address-of-packed-member -Waddress-of-packed-member: YES
Compiler for C supports arguments -Wno-packed-not-aligned -Wpacked-not-aligned: YES
Compiler for C++ supports arguments -Wno-packed-not-aligned -Wpacked-not-aligned: YES
Compiler for C supports arguments -Wno-missing-field-initializers -Wmissing-field-initializers: YES
Compiler for C++ supports arguments -Wno-missing-field-initializers -Wmissing-field-initializers: YES
Fetching value of define "SSE4_2" : 1
Fetching value of define "AES" : 1
Fetching value of define "PCLMUL" : 1
Fetching value of define "AVX" : 1
Fetching value of define "AVX2" : 1
Fetching value of define "AVX512F" :
Fetching value of define "RDRND" : 1
Fetching value of define "RDSEED" : 1
Fetching value of define "AVX512F" : (cached)
Fetching value of define "AVX512DQ" :
Message: *** IPv4 dump output disabled
Message: *** TCP support disabled
Message: *** Enable support for punting to Linux kernel stack
Message: *** TCP dump output disabled
Configuring cne_build_config.h using configuration
Fetching value of define "AVX2" : 1 (cached)
Fetching value of define "AVX512F" : (cached)
Fetching value of define "AVX512VL" :
Fetching value of define "AVX512CD" :
Fetching value of define "AVX512BW" :
Compiler for C supports arguments -mavx512f -mavx512vl -mavx512cd -mavx512bw: YES
Compiler for C supports arguments -mavx512f -mavx512dq: YES
Message: *** have multiple avx512f and avx512dq
Compiler for C supports arguments -mavx512bw: YES
Compiler for C supports arguments -mavx512bw: YES
Program python3 found: YES (/usr/bin/python3)
Program doxygen found: YES (/usr/bin/doxygen)
Program generate_doxygen.sh found: YES (/home/canopus/cndp/doc/api/generate_doxygen.sh)
Program generate_examples.sh found: YES (/home/canopus/cndp/doc/api/generate_examples.sh)
Program doxy-html-custom.sh found: YES (/home/canopus/cndp/doc/api/doxy-html-custom.sh)
Configuring doxy-api.conf using configuration
Program sphinx-build found: YES (/usr/bin/sphinx-build)
Message: Fuzz tests require clang. Fuzz tests will not be built.
Message: **** Missing Linux source code, not building afxdp_user example:/usr/src/linux-source-5.13.0
Library quicly found: NO
Message: **** Quicly not found at /work/projects/intel/networking.dataplane/quicly
Update meson_options.txt file.
The quic-echo example cannot be built.
Message: DLB test build missing library
Program tools/mklib.sh found: YES (/home/canopus/cndp/tools/mklib.sh)
Program cp found: YES (/bin/cp)
Program cp found: YES (/bin/cp)
Message: >>> Create pkg-config file
Message: <<< Done pkg-config file
Build targets in project: 69

Option default_library is: static [default: shared]
Found ninja-1.10.0 at /usr/bin/ninja
ninja: Entering directory /home/canopus/cndp/builddir' [75/309] Linking target lib/core/pmds/net/af_xdp/libpmd_af_xdp.so. FAILED: lib/core/pmds/net/af_xdp/libpmd_af_xdp.so cc -o lib/core/pmds/net/af_xdp/libpmd_af_xdp.so 'lib/core/pmds/net/af_xdp/3c1feb5@@pmd_af_xdp@sha/pmd_af_xdp.c.o' -Wl,--as-needed -Wl,--no-undefined -Wl,-O1 -shared -fPIC -Wl,--start-group -Wl,-soname,libpmd_af_xdp.so -Wl,--no-as-needed -pthread -lm -ldl -lbsd -ljson-c -lpcap -lnl-3 -lnl-cli-3 -lnl-route-3 lib/core/osal/libcne_osal.a lib/core/log/libcne_log.a lib/core/mmap/libcne_mmap.a lib/core/cne/libcne_cne.a lib/core/kvargs/libcne_kvargs.a lib/core/pktdev/libcne_pktdev.a lib/core/mempool/libcne_mempool.a lib/core/ring/libcne_ring.a lib/core/pktmbuf/libcne_pktmbuf.a lib/common/uds/libcne_uds.a lib/core/xskdev/libcne_xskdev.a /usr/lib/libbpf.a /usr/lib/x86_64-linux-gnu/libelf.a /usr/lib/x86_64-linux-gnu/libz.a -Wl,--end-group '-Wl,-rpath,$ORIGIN/../../../osal:$ORIGIN/../../../log:$ORIGIN/../../../mmap:$ORIGIN/../../../cne:$ORIGIN/../../../kvargs:$ORIGIN/../../../pktdev:$ORIGIN/../../../mempool:$ORIGIN/../../../ring:$ORIGIN/../../../pktmbuf:$ORIGIN/../../../../common/uds:$ORIGIN/../../../xskdev' -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/osal -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/log -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/mmap -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/cne -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/kvargs -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/pktdev -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/mempool -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/ring -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/pktmbuf -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/common/uds -Wl,-rpath-link,/home/canopus/cndp/builddir/lib/core/xskdev /usr/bin/ld: /usr/lib/x86_64-linux-gnu/libelf.a(elf_error.o): relocation R_X86_64_TPOFF32 against global_error' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: /usr/lib/libbpf.a(libbpf.o): relocation R_X86_64_PC32 against symbol `stderr@@GLIBC_2.2.5' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status
[84/309] Compiling C object 'lib/usr/clib/acl/ee741e7@@avx512_tmp@sta/acl_run_avx512.c.o'.
ninja: build stopped: subcommand failed.
make: *** [Makefile:46: rebuild] Error 1

Difficulty: Medium to Hard depends are understanding of GitHub Actions.

Potential enhancement idea for future work - it seems like it should be possible to cache part of the docker build layers to improve build times using the GitHub Actions cache system. The docker build is far slower than all the other build steps combined.

Tricky part is to do so in a way that doesn't eventually fill up the actions cache. We were thinking about trying to do this but ran out of time. Here's a decent article with a possible approach to this - https://evilmartians.com/chronicles/build-images-on-github-actions-with-docker-layer-caching

Maintainers, feel free to close or comment on this if you would be interested in this in the future.

Difficulty: Easy to Medium

Update the CNDP Ansible playbook to make sure it support for Fedora installs.

The playbook needs to be updated to support libbpf and libxdp as the new libxdp contains the XSK APIs which were deprecated in libbpf , but libbpf is still required.

Difficulty: Medium

The l3fwd_fib_lookup function used in the RX loop in the cndpfwd _l3fwd_test case currently performs a single FIB lookup each time it's called, using the cne_fib_lookup_bulk function.

As its name suggests, cne_fib_lookup_bulk is designed to perform n lookups at a time. In its current usage, we call it repeatedly with an n of 1. A more idiomatic use of this function would be to construct the array of IPs to look up in the _l3fwd_test RX loop, perform the lookups in bulk with a single call to cne_fib_lookup_bulk, then iterate on the result set to modify & TX the packets.

I appears PR #177 keeping the FQ full causes the performance to drop on nightly performance tests.

It maybe because the pktmbuf_t alloc/free calls, but we need to figure it out and rework the xskdev some to regain performance. The cndpfwd example appeared to work correctly and it appeared to fix the Go application performance when using rx_burst() requests larger than 64 mbufs. Need to make sure the performance drop is real in the nightly builds to make sure it is not a false positive or something with the configuration of the performance setup.

I followed the CNDP instructions here: https://github.com/CloudNativeDataPlane/cndp/blob/main/INSTALL.md but had an issue with the libbpf step:

Install libbpf from source

https_proxy="https://user:[email protected]:port" git clone https://github.com/libbpf/libbpf.git
cd libbpf
git checkout ...
make -C src
sudo make -C src install

There are 2 issues here. First I don't think the https_proxy needs to be there. Secondly this gives a version of libbpf (0.8.0/??) which seems incompatible with cndp:

$ make
"/work/cndp/tools/cne-build.sh" build
Build environment variables and paths:
  CNE_SDK_DIR     : /work/cndp
  CNE_TARGET_DIR  : usr/local
  CNE_BUILD_DIR   : /work/cndp/builddir
  CNE_DEST_DIR    : /work/cndp
  PKG_CONFIG_PATH : /usr/lib64/pkgconfig
  build_path      : /work/cndp/builddir
  target_path     : /work/cndp/usr/local

>>> Release build in '/work/cndp/builddir'
>>> Ninja build in '/work/cndp/builddir' buildtype='release'
usage: meson [-h] {setup,configure,dist,install,introspect,init,test,wrap,subprojects,help,rewrite,compile,devenv} ...
meson: error: unrecognized arguments: 
ninja: Entering directory `/work/cndp/builddir'
[10/54] Compiling C object lib/core/xskdev/libcne_xskdev.so.p/xskdev.c.o
FAILED: lib/core/xskdev/libcne_xskdev.so.p/xskdev.c.o 
cc -Ilib/core/xskdev/libcne_xskdev.so.p -Ilib/core/xskdev -I../lib/core/xskdev -Ilib/include -I../lib/include -Ilib/core/osal -I../lib/core/osal -Ilib/core/log -I../lib/core/log -I. -I.. -Ilib/core/cne -I../lib/core/cne -Ilib/common/uds -I../lib/common/uds -Ilib/core/mmap -I../lib/core/mmap -Ilib/core/mempool -I../lib/core/mempool -Ilib/core/pktmbuf -I../lib/core/pktmbuf -I/usr/include/libnl3 -fdiagnostics-color=always -D_FILE_OFFSET_BITS=64 -Wall -Winvalid-pch -Wextra -Wpedantic -Werror -O3 -g -Wno-pedantic -Wcast-qual -Wdeprecated -Wformat-nonliteral -Wformat-security -Wmissing-declarations -Wmissing-prototypes -Wnested-externs -Wold-style-definition -Wpointer-arith -Wsign-compare -Wstrict-prototypes -Wundef -Wwrite-strings -Wno-address-of-packed-member -Wno-packed-not-aligned -Wno-missing-field-initializers -D_GNU_SOURCE -include cne_build_config.h -march=native -DCC_AVX2_SUPPORT -fPIC -MD -MQ lib/core/xskdev/libcne_xskdev.so.p/xskdev.c.o -MF lib/core/xskdev/libcne_xskdev.so.p/xskdev.c.o.d -o lib/core/xskdev/libcne_xskdev.so.p/xskdev.c.o -c ../lib/core/xskdev/xskdev.c
../lib/core/xskdev/xskdev.c: In function 'configure_busy_poll':
../lib/core/xskdev/xskdev.c:122:5: error: 'xsk_socket__fd' is deprecated: libbpf v0.7+: AF_XDP support deprecated and moved to libxdp [-Werror=deprecated-declarations]
  122 |     int fd            = xsk_socket__fd(rxq->xsk);
      |     ^~~
In file included from ../lib/core/xskdev/xskdev.h:11,
                 from ../lib/core/xskdev/xskdev.c:27:
/usr/include/bpf/xsk.h:257:5: note: declared here
  257 | int xsk_socket__fd(const struct xsk_socket *xsk);
      |     ^~~~~~~~~~~~~~

This is a useful hint buried there but it isn't obvious:

 error: 'xsk_socket__fd' is deprecated: libbpf v0.7+: AF_XDP

In order to get cndp to compile I had to use libbpf 0.6.1. I don't know if that is the preferred version but it worked.

cd libbfp
git checkout v0.6.1
make -C src && sudo make -C src install

In #223, we need some more examples to show the usage of CNDP Go Binding layer.

Here we track a detail example of packet distributor, it performs the distribution of packets that are received on an rx queue to different cores to process, and send to an tx queue to transmit.

Creating a chain of virtual network functions (VNFs) is an intriguing topic. Is this something that Kubernetes stack would take care of?

Are there any methods that could be used to achieve this without compromising performance but in a modular way and not involving Kubernetes?

IMHO this framework does have a great potential to be used outside of Kubernetes.

Difficulty: Hard

Modify CNDP mempool code to allow multiple processes to share a single mempool across processes. CNDP is a single process model application type and we need to be able to share buffers between processes without needing to share the entire process memory space which is what DPDK does today.

Sharing memory between processes is normally done with mmap() API to create a sharable region of memory. When sharing memory between processes you can not assume the virtual address of the memory space is the same between processes. That means memory pointers are not allowed in the shared memory region. This means the pointer currently being used need to be convert to offsets into the memory region.

The implementation should rely on Linux IPC and existing shared memory infrastructure wherever possible.

It should be possible to send and receive packets using AF_XDP sockets which use a shared UMEM between two or more processes.

Following the instructions for the rust APIs through the readme allows for a successful build of those APIs but the unit tests seem to be failing when running the ./run_cne_test.sh script :

running 8 tests
test tests::test_register ... ok
test tests::test_load_config ... FAILED
test tests::test_get_port ... FAILED
test tests::test_get_port_details ... FAILED
test tests::test_cne_instance ... FAILED
test tests::test_loopback ... FAILED
test tests::test_rx_burst ... FAILED
test tests::test_tx_burst ... FAILED

failures:

---- tests::test_load_config stdout ----
thread 'tests::test_load_config' panicked at 'assertion failed: cfg.is_ok()', apis/cne/src/lib.rs:80:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- tests::test_get_port stdout ----
thread 'tests::test_get_port' panicked at 'assertion failed: ret.is_ok()', apis/cne/src/lib.rs:118:9

---- tests::test_get_port_details stdout ----
thread 'tests::test_get_port_details' panicked at 'assertion failed: ret.is_ok()', apis/cne/src/lib.rs:152:9

---- tests::test_cne_instance stdout ----
thread 'tests::test_cne_instance' panicked at 'assertion failed: ret.is_ok()', apis/cne/src/lib.rs:99:9

---- tests::test_loopback stdout ----
thread 'tests::test_loopback' panicked at 'assertion failed: ret.is_ok()', apis/cne/src/lib.rs:282:9

---- tests::test_rx_burst stdout ----
thread 'tests::test_rx_burst' panicked at 'assertion failed: ret.is_ok()', apis/cne/src/lib.rs:180:9

---- tests::test_tx_burst stdout ----
thread 'tests::test_tx_burst' panicked at 'assertion failed: ret.is_ok()', apis/cne/src/lib.rs:233:9


failures:
    tests::test_cne_instance
    tests::test_get_port
    tests::test_get_port_details
    tests::test_load_config
    tests::test_loopback
    tests::test_rx_burst
    tests::test_tx_burst

test result: FAILED. 1 passed; 7 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s

error: test failed, to rerun pass `-p cndp-cne --lib`

The AF_XDP device plugin only sets the AFXDP_DEVICES environment variable.

The script needs to be updated to set AFXDP_COPY_MODE and LIST_OF_QIDS appropriately

Initial work was pushed in this PR

Just with a simple go program main.go.txt that loops receiving packets, I am getting zero length packets (even at a low rate of 100pps, with all > 124bytes).
This is on an interface with just 1 channel configured, running on ubuntu (setup with ansible scripts).
If I up the bufcnt to 8 I don't see these anymore.

Difficulty: Medium to hard if you need to learn Rust.

Create a new Rust example using the Rust binding layer APIs. Need to create or port a Rust application using the binding layer. The application needs to do more than a simple forwarding application. The Rust binding layer has two layers one is a low level set API, which is created by the Rust bindgen application to expose all of the CNDP APIs to Rust. The second set of APIs is a higher level of APIs and these are the ones the application needs to be using. Adding this application could require updating or creating new high level APIs.

The rust build inside the fedora oci image is currently failing
reproduce with: $ make oci-fed-image

#0 87.91   --- stderr
#0 87.91   thread 'main' panicked at 'Non floating-type complex? Type(_Complex _Float16, kind: Complex, cconv: 100, decl: Cursor( kind: NoDeclFound, loc: builtin definitions, usr: None), canon: Cursor( kind: NoDeclFound, loc: builtin definitions, usr: None)), Type(_Float16, kind: Float16, cconv: 100, decl: Cursor( kind: NoDeclFound, loc: builtin definitions, usr: None), canon: Cursor( kind: NoDeclFound, loc: builtin definitions, usr: None))', /root/.cargo/registry/src/github.com-1ecc6299db9ec823/bindgen-0.59.2/src/ir/context.rs:1992:26
#0 87.91   note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
#0 87.91 warning: build failed, waiting for other jobs to finish...
#0 88.31 error: failed to compile `fwd v0.1.0 (/cndp/lang/rs/examples/fwd)`, intermediate artifacts can be found at `/cndp/lang/rs/target`
------
Dockerfile:76
--------------------
  74 |     # 'fwd' binary will be installed in /cndp/usr/local/bin
  75 |     WORKDIR /cndp/lang/rs
  76 | >>> RUN CNDP_INSTALL_PATH="/cndp" cargo install --root /cndp/usr/local/ --path examples/fwd
  77 |     
  78 |     WORKDIR /cndp
--------------------
ERROR: failed to solve: process "/bin/sh -c CNDP_INSTALL_PATH=\"/cndp\" cargo install --root /cndp/usr/local/ --path examples/fwd" did not complete successfully: exit code: 101
make: *** [Makefile:118: oci-fed-image] Error 1

Remove references to the ClusterType flag based on the latest changes in: intel/afxdp-plugins-for-kubernetes#46

I have posted question for stack overflow for which I need theanswer.

https://stackoverflow.com/questions/76991150/port-rx-drop-in-intel-corporation-ethernet-controller-xl710-for-40gbe-qsfp?noredirect=1#comment135735741_76991150

I created a cndp forwarding application. And tested it on 40gbps NIC i found that the performance decreases when both ingress and egress traffic on the nic.

With only ingress traffic on the NIC i am able to get 40gbps but when both ingress and egress traffic on NIC is there i get at max 23gbps.
I don't know the reason why it reduces so much.

"umem0": {
"bufcnt": 400,
"bufsz": 2,
"mtype": "2MB",
"regions": [
400
],
"rxdesc": 4,
"txdesc": 4,
"cache": 256,
"description": "UMEM Description 0"
},

This is my umem configuration.

Any insights will be very helpful.

Currently if we are using the uds we are setting the LPORT_UNPRIVILEGED in the example app rather than the underlying library.

            if (f->xdp_uds) {
                pcfg.xsk_uds = f->xdp_uds;
                lport->flags |= LPORT_UNPRIVILEGED;
            }

It's better if the lport flag is configured in the core library rather than the example (more under the hood rather than having user responsibility to set it)... it might be trickier through the jcfg libraries but less prone to error in the longer term

Currently with the ip4_forward node we call fib_info_lookup with the arp fib table. We need to use the route fib table in conjunction with the arp table.

fi          = cnet->arp_finfo;
....
 if (unlikely(fib_info_lookup(fi, ip4, (void **)arp, 4) > 0)) {
 ...

with fib_info_lookup if we rely so much on the arp FIB table and not route FIB the ip4_forward node won't know what to do with the traffic that doesn't have an arp entry and the node will assign it to the arp_request node.... the suggestion is to do an arp lookup first, if that fails - try to find the next hop info from the route fib... we may then need to do an arp lookup on that next hop...

Difficulty: Medium

Add L3 forwarding to cndpfwd using RIB/FIB APIs.

Hi Folks
I recently started looking through the Rust language bindings for CNDP and ran into a few issues. Following the readme things don't work out of the box. In addition the directory structure is completely confusing (there's no real structure to what's an example vs what's the set of libraries I can use as part of my Rust application). Also I believe the naming should be libcndp-sys for the directory that contains the language bindings (cne is only a small library set of what CNDP has to offer).

The current dir structure is shown below:

.
├── README.md
├── bindings
│   ├── cne
│   │   ├── Cargo.toml
│   │   ├── README.md
│   │   ├── examples
│   │   │   └── loopback
│   │   │       ├── Cargo.toml
│   │   │       ├── run.sh
│   │   │       └── src
│   │   │           └── main.rs
│   │   ├── fwd.jsonc
│   │   ├── run_test.sh
│   │   └── src
│   │       ├── config.rs
│   │       ├── error.rs
│   │       ├── instance.rs
│   │       ├── lib.rs
│   │       ├── packet.rs
│   │       ├── port.rs
│   │       └── util.rs
│   ├── cne-sys
│   │   ├── Cargo.toml
│   │   ├── README.md
│   │   ├── build.rs
│   │   └── src
│   │       ├── bindings
│   │       │   ├── bindings.c
│   │       │   ├── bindings.h
│   │       │   ├── meson.build
│   │       │   └── wrapper.h
│   │       └── lib.rs
│   └── examples
│       └── echo_server
│           ├── Cargo.toml
│           ├── fwd.jsonc
│           ├── run.sh
│           └── src
│               └── main.rs
├── helloworld
│   └── main.rs
├── mmap
│   ├── Cargo.toml
│   └── src
│       └── lib.rs
├── mmap-cndp
│   ├── Cargo.toml
│   ├── build.rs
│   ├── src
│   │   └── lib.rs
│   └── wrapper.h
├── pktfwd
│   ├── Cargo.toml
│   ├── build.rs
│   ├── fwd.jsonc
│   ├── jcfg_parse
│   │   ├── fwd.h
│   │   ├── meson.build
│   │   ├── parse-jsonc.c
│   │   ├── rust_helper.c
│   │   ├── rust_helper.h
│   │   └── stats.c
│   ├── runcmd.sh
│   ├── src
│   │   ├── cndp.rs
│   │   ├── main.rs
│   │   ├── packet.rs
│   │   └── util.rs
│   └── wrapper.h
├── ring_test
│   ├── Cargo.toml
│   ├── build.rs
│   ├── src
│   │   ├── lib.rs
│   │   └── main.rs
│   └── wrapper.h
└── wireguard
    └── patch
        ├── 0001-Integrate-CNDP-Cloud-Native-Data-Plane-with-Wireguar.patch
        ├── 0002-Rename-variable-private-to-priv_-to-fix-build-error.patch
        └── 0003-Remove-extra-argument-from-pktmbuf_dump.patch

Finally - there's only a handful of APIs that have been converted to Rust and the naming is separate to the C libraries - so tracking what's where is going to be difficult.

Is there a plan to cleanup the Rust implementation and port more libraries?

Hello, I'm interested in XDP and finding a new solution to enhance XDP with hardware support offloading.
I run "examples/cndpfwd" but the current version seems to only run in XDP native mode.
Would you show me some way I can run and offload XDP on Netronome’s Agilio Smart NIC.

Thank you.

We have network interfaces that are reaching the maximum device name length.

(pktdev_allocate : 84) ERR: Ethernet device name is too long
(init_lport : 209) ERR: pktdev_allocate(enp24s0f0np0:10, enp24s0f0np0) failed
(pmd_af_xdp_probe : 248) ERR: Failed to init lport
(pktdev_port_setup : 50) ERR: driver probe(enp24s0f0np0:net_af_xdp) failed
2023/11/01 02:32:01 cndp init failed: pktdev_port_setup() failed for lport enp24s0f0np0:10

and

cndp init failed: pktdev_port_setup() failed for lport enp152s0f0np0:0

Is it possible to increase the PKTDEV_NAME_MAX_LEN value from 16?

Difficulty: Easy

Enhance the Go binding test modules to improve coverage and possible show performance of the APIs. The Go binding module and/or the functional tests in the testing module.

lang/go/bindings/cne$ ./run_test
=== RUN TestStarting
--- PASS: TestStarting (0.33s)
=== RUN TestGetChannel
=== RUN TestGetChannel/TestLoopback
=== RUN TestGetChannel/TestLoopback#01
--- PASS: TestGetChannel (0.00s)
--- PASS: TestGetChannel/TestLoopback (0.00s)
--- PASS: TestGetChannel/TestLoopback#01 (0.00s)
=== RUN TestRxBurst
=== RUN TestRxBurst/TestRxBurst
cne.test: ../lib/core/cne/cne.c:28: cne_id: Assertion `per_thread__cne != NULL' failed.
SIGABRT: abort
PC=0x7fdce9f0000b m=3 sigcode=18446744073709551610
signal arrived during cgo execution

The same thing happens when I run OpenWIthConfig() in my own code.

Currently CNDP applications using xskdev.h header needs to include libbpf-dev as dependency because xskdev.h includes <bpf/xsk.h>. Is it possible to refactor xskdev_info_t structure in xskdev.h such a way that we can include “xskdev_xxx” data types in some private structure defined in an internal header file like xskdev_internal.h and then have a member variable like void *priv in xskdev_info_t structure.

File: xskdev_internal.h

#include <bpf/xsk.h>

typedef struct xskdev_info_internal {
    xskdev_rxq_t rxq;              /**< RX queue */
    xskdev_txq_t txq;              /**< TX queue */
    xskdev_get_mbuf_addr_tx_t     __get_mbuf_addr_tx;         /**< Internal function to set the mbuf address on tx */
    xskdev_get_mbuf_rx_t           __get_mbuf_rx;                   /**< Internal function to get the mbuf address on rx */
    xskdev_pull_cq_addr_t          __pull_cq_addr;                  /**< Internal function to pull the complete queue */
}

File: xskdev.h

typedef struct xskdev_info {
      …..

      void *priv; /**  >struct xskdev_info_internal > **/
       …..
}

File: xskdev.c

#include  xskdev_internal.h
#include xskdev.h

xskdev_info  xskdev;
struct xskdev_info_internal *xskdev_internal = (struct xskdev_info_internal *) xskdev->priv;

Difficulty: Easy

Enhance the test-cne application to improve an existing test or add a new one for testing a library not previously supported i.e., event/osal/txbuff/... library. Some of these tests could also have a performance test to show the performance of the APIs.

when cndp can support dev bond feature?
for example dpdk bond poll mode.
http://doc.dpdk.org/guides/prog_guide/link_bonding_poll_mode_drv_lib.html

Difficulty: Easy to Medium

I noticed that the error messages for when calloc returns NULL use different strings in different parts of the codebase.

To see an example of what I'm talking about, try this against the root of the repo:
for s in "Unable" "Failed" "Allocation"; do grep -rn -A3 calloc | grep $s; done

There are probably other contexts where a similar thing happens. I think the most thorough fix for this would probably look like enumerating those contexts and perhaps adding some standards to the coding style guide.

Difficulty: Easy to Medium

Some additional cleanup is needed to ensure that malloc'ed/mmap'd memory is freed when we return in error during argument parsing in: https://github.com/CloudNativeDataPlane/cndp/blob/main/examples/cndpfwd/parse-args.c