[BUG] DHCP Ethernet crashes after repeated switching of network interface [OPTA/Portenta H7] #891
Comments
|
@Channel59 thank you for submitting this problem.
|
|
Dear @maidnl, Thank you very much for taking the time to investigate this problem. I agree with your remarks regarding the replication of the problem. I have tested your patch and it seems like the issue is partly resolved. However, the opta can still become unresponsive quite easily, using e.g. this code: #include <Arduino.h>
#include "EthernetInterface.h"
#include "mbed.h"
auto net = EthernetInterface::get_default_instance();
void setup(){
net->set_blocking(true);
net->set_dhcp(true);
net->connect();
}
void loop() {
rtos::ThisThread::sleep_for(1000);
}and by simply using this commands (or unplugging the ethernet cable for 3 seconds and re-plugging it): sudo ifconfig eth0 down && sleep 3 && sudo ifconfig eth0 upThis is already enough to break it, however re-running this will also restore the connecition With static IP there is no issue it seems. or with this more elaborate script, including your wait time: #!/bin/bash
opta=192.168.0.100
eth=eth0
for i in {1..2000}
do
for j in {1..5}
do
t_down=$((500 + $RANDOM % 5000))
t_up=$((1000 + $RANDOM % 2000))
echo -ne "_"
ifconfig $eth down
./usleep $(($t_down * 1000))
echo -ne "\b-"
ifconfig $eth up
./usleep $(($t_up * 1000))
echo -ne "\b."
done
if (( $t_up<3000 )); then
t_wait=$((3000 - t_up))
./usleep $(($t_wait * 1000))
fi
sleep 2
if ping -c 1 $opta &> /dev/null
then
echo -n "|"
else
echo "X"
exit 1
fi
done(Use e.g. this for usleep: https://github.com/pklaus/usleep-binary) In this case, another ethernet interface down/up from the host seems to resolve the problem and the opta starts sending DHCP requests again and obtains an ip. It seems there is still something stange going on with the DHCP. As a side-note, that may or may not be related, I have done some more testing of the Opta in the context of ethernet disconnect/reconnects, and have found an issue that is seems to be related UDP multicast (see:#904). I assume the DHCP uses somewhere under the hood also some UDP multicast socket, and thus these issues might be related / the same. |
|
@Channel59 Thank you for test it, we hoped that this would fix this problem. This issue is then closes thanks to commit 97304bffff1852de6ebf10ccbb941f431b70b16 |
I experience some issues with the network interface of the OPTA (probably also the portenta H7 suffers from this problem).
I am not sure if only the OPTA is affected, and whether or not this issue is OPTA related, arduinoCore-mbed related or really mbed-os related, or a combination of those.
Setup:
PC with multiple eth interface, Opta directly connected to
eth0. The PC has a static IP and runs a DHCP server. ArduinoCore-mbed version: 4.0.10Issue:
If from the PC the interface
eth0is brought up and down multiple times, the OPTA sees this and the LAN light turn off when the interface is down and turn back on when the interface is up.If the OPTA is configured to use a static IP this gives no issues. However, when DHCP is enabled, the OPTA seems to be unable to reconnect to the network. It cannot be pinged either. Also, the network stack of the OPTA does not seem to notice this, because the state will actually become
NSAPI_STATUS_GLOBAL_UP, but it does not respond to pings.The same behavior can also be reproduced by repeatedly pulling the inserting the ethernet cable in the RJ-45 port.
With a static IP this seems not to be an issue.
Recovering from this state
It seems that the only way to recover from this state is by resetting the OPTA, which is undesirable. It seems that removing and re-inserting the ethernet cable can also recover the OPTA from this state.
Code on the OPTA:
This issue happened when using Arduino_ConnectionHandler, just the PortentaEthernet class from the arduinocore, and it also happen when I just use the mbed EthernetInterface.
The simplest way to reproduce the problem is with this code:
The script to break the opta:
Outputs:
First we ping the opta to check if it has gotten a DHCP lease:
This seems normal.
Then we run the
break-opta.shscript.The opta cannot be reached anymore.
If we look at the serial output from the opta, it seems that everything is fine:
A reset of the opta seems to be the only way to recover from this state.
Attempts to recover without resetting
Several things I have tried (behind a button push in the loop() function):
Furthermore, I tried to build the project with the lwip headers included, by adding these values to
variants/OPTA/includes.txt:This allowed me to access the functions, calling them after the button push. None of these seemed to be able to recover the network connection.
None of these things seem to be able to restore the network status.
Does anyone have an idea as to how to fix this?
The text was updated successfully, but these errors were encountered: