| Age | Commit message (Collapse) | Author |
|---|
|
The tests create the veth pair with 'ip link add ... type veth peer name
...' and let the kernel pick random hardware addresses, then start
accel-pppd right away. On distributions where systemd-udevd rewrites the
address of a freshly created veth (MACAddressPolicy in its .link rules,
seen on Ubuntu 22.04) this is a race: accel-pppd reads the address of the
interface it serves once, at startup, and when udev wins it advertises an
address the interface no longer has.
Discovery still completes - the client copies the advertised address into
its kernel PPPoE session as the peer address - but from then on the two
ends disagree about it, no PPP frame is exchanged in either direction, and
the session dies on LCP timeouts. Every PPPoE session test can fail this
way; test_pppoe_ccp_ipcp_race failed 8 times out of 12 runs in one qemu
Ubuntu-22.04 job.
Create the pair with fixed, locally administered addresses derived from
the name the pair already gets, so udev has nothing to change. The same
job passes 12 out of 12 with this.
This only removes the trigger in the tests. accel-pppd not following the
address of the interface it serves is tracked in #363 and needs a fix of
its own.
|
|
Three tests around the interfaces the ipoe module creates per session:
- kill accel-pppd with SIGKILL while a dhcp session is up, start it
again and check that the interfaces left behind by the killed
instance are gone,
- remove a session interface with 'ip link del',
- check that 'ip link add ... type ipoe' is refused, since a device
made through rtnetlink would have none of the private state that
IPOE_CMD_CREATE sets up.
They carry the ipoe_driver marker and sit next to the existing ipoe
tests, so the workflows run them in the steps that follow the insmod of
the module. No workflow change is needed.
The [modules] section has to list connlimit and radius before ipoe:
libipoe.so refers to symbols of both, and with a strict dynamic linker
loading it on its own fails with a relocation error instead of a missing
feature. The tests assert that accel-pppd came up, so that this kind of
misconfiguration is not reported as an unrelated cli connection failure.
Restarting accel-pppd needs no explicit synchronisation:
triton_load_modules() runs every DEFINE_INIT() before triton_run()
starts the threads that serve the cli, so by the time accel-cmd 'show
version' is answered the flush registered at DEFINE_INIT(19) has already
run.
Interfaces are compared by ifindex and not by name, because dhclient may
get a new session in the meantime and the fresh interface would reuse
the ipoe0 name. The check that nothing removes the interfaces while no
accel-pppd is running is only printed, not asserted, so that the module
is free to start doing it on its own later on.
|
|
|
|
This commit adds tests (using python3 pytest framework):
1. Test basic accel-cmd commands (show version, show stat, etc)
2. Test ipoe shared session up (dhcpv4) without radius
3. Test pppoe discovery (without PADO delay)
4. Test pppoe discovery (without PADO delay)
5. Test pppoe session up (ipv4) without radius
6. Test vlan creation using vlan-mon (pppoe)
These tests require external utils. Please read tests/README.md how to setup
environment, how to run the tests and how to generate coverage report
Also, run-tests.yml contains step-by-step instruction how to run the tests
Signed-off-by: Sergey V. Lobanov <sergey@lobanov.in>
|
