To decide whether we should use the compatibility mode or the normal
mode with a peer, we use the handshake messages as a signaling channel.
If we receive the expected ECN bits, it most likely means they're
running a compatible version.
Signed-off-by: Florent Daigniere <nextgens@freenetproject.org>
Like many, I am using WiFi a lot and often on congested networks.
Without this, Wireguard strips the DSCP bits, preventing WME from
kicking in and improving the audio/video experience.
Yes, it's technically an information leak. Who cares? It's not like if
traffic analysis based on packet sizes or timings wasn't a thing.
This is the first patch of the serie, more work has to happen on ECN
Signed-off-by: Florent Daigniere <nextgens@freenetproject.org>
If the queues are full, we drop the present packet, which is better for
network traffic flow. Also, we try to fix up the memory leaks with not
putting buffers from our shared pool.
We reorganize this into also doing so on sending keepalives itself,
which means the state machine is much more consistent, even if this was
already implied.
Kernel module commit 30290ef1d2581a3e6ee8ffcdb05d580cfba976be
Now listen for flushNonceQueue signal in outer select during
the RoutineNonce routine. This is needed to handle the edge case
where the queue is flushed, but no packets are in the nonce queue.
Since the signal has capacity 1 this signal will remain and potentially
flush the queue at a later time, with packets meant for transmission.
1. Removed remaining signals from peer struct
2. Made needAnotherKeepalive local
3. Removed environment check from warning text (annoying when debugging)
Flushing queues on exit is sort of a partial solution, but this could be
better. Really what we want is for no more packets to be enqueued after
isUp is set to false.
A premature waitgroup .Done resulted in reading from closed channel.
This caused a nil-pointer deref & crash.
Added additional debugging when closing routines.
