Already merged into master branch:
https://github.com/accel-ppp/accel-ppp/pull/115

Hi Leonardo

Use RPS or RFS https://www.kernel.org/doc/html/latest/networking/scaling.html?highlight=rps

Okey, so in that case, how do you deal with it?

this is not related to accel, because accel-ppp is only control-plane for pppoe sessions so slow performance of pppoe is related to kernel, not to accel-ppp

Only difference is type of accel session. ipoe vs pppoe. So this is not related to accel?

This question is not related to accel, traffic forwarding is done by kernel, not by accel-ppp so please contact netdev (https://lore.kernel.org/netdev/ ) if you want to discuss it

Hi @Dimka88, I have implemented this feature for personal needs, but I would like to contribute with the code if possible. Please let me know how can I upload the changes so you can review. Should I do this on Github?

Summarize:
Issue appears with veth and vlan_mon only on 4.19. If use bridge instead, all will work properly. As Deb10 support expired around 1 year we make decision to delete tests on Deb10 from github test

I can confirm that kernel 5.10.0-0.deb10.24-amd64 is fixing issue on Debian 10.

It is not fixed in actual version. It is required to fix here: https://github.com/accel-ppp/accel-ppp/blob/2b865db72bc2ddc6411950d72f1c23e8ef115b8a/accel-pppd/ppp/ipcp_opt_ipaddr.c#L100 , https://github.com/accel-ppp/accel-ppp/blob/2b865db72bc2ddc6411950d72f1c23e8ef115b8a/accel-pppd/ppp/ipv6cp_opt_intfid.c#L209 , https://github.com/accel-ppp/accel-ppp/blob/2b865db72bc2ddc6411950d72f1c23e8ef115b8a/accel-pppd/ctrl/ipoe/ipoe.c#L668

Attached the patch i made for the version 1.12.0-202-g28fe4de. It will use Calling Number/Called Number as RADIUS Calling-Station-Id/Called-Station-Id in case they are present in L2TP ICRQ, otherwise old beahvior is used (LAC/LNS IP addresses).

UPD:// CMake Error looks pretty strange. It get info by the command git describe --tags --always
From your repo it return 6e5f998 because you clone only single branch.

root@debian11:/opt/accel-ppp/build# git describe --tags --always
6e5f998

So, maybe we have to improve generate version functionality also

Hi @v.huti , I see some cmake Errors

root@debian11:/opt/accel-ppp/build# cmake -DBUILD_IPOE_DRIVER=TRUE -DBUILD_VLAN_MON_DRIVER=TRUE -DCMAKE_INSTALL_PREFIX=/usr -DKDIR=/usr/src/linux-headers-`uname -r` -DLUA=TRUE -DCPACK_TYPE=Debian11 ..
-- The C compiler identification is GNU 10.2.1
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- 'x86_64'
CMake Error at cmake/cpack.cmake:5 (list):
  list index: 1 out of range (-1, 0)
Call Stack (most recent call first):
  CMakeLists.txt:49 (include)

Hi @Dimka88! The warnings are fixed in the following pull request:
https://github.com/accel-ppp/accel-ppp/pull/64
From my understanding, the situation is next:

In general, assuming anything about data alignment is a bad practice that leads to very obscure errors. 
Taking the address of a packed member is dangerous since the reduced alignment of the pointee is lost.
If the pointer value is dereferenced, this can lead to memory alignment faults in some architectures.

I think it is not implemented yet, but is is good feature request