No internet connection after upgrade from April 20 snapshot to May 19

motific

There was an issue with routes in a snapshot a couple of weeks back, are you getting any error messages in the console? I couldn’t reproduce the problem after resetting default routes, if you can then that could be useful info. (Check your IP6 routes too if you have them.)

w0w

IPv6 is disabled (not enabled). Do you use PPPoE as WAN?

UPDATE: So, I've found that we have new option in gateway options. This one is called
Use non-local gateway
"Use non-local gateway through interface specific route. This will allow use of a gateway outside of this interface's subnet. This is usually indicative of a configuration error, but is required for some scenarios."

After enabling this option and setting gateway as default — traffic enabled and reached Internet and vice versa. The question remaining is it bug or feature. Does it checks that gateway is not from the same subnet and therefore does not make it default?
My ISP gateway is really not from the same subnet as DHCP PPPoE IP
My IP is 84.x2.5x.12x, ISP gateway is 212.7.29.236 and they both received and configured automagically via PPPoE

jimp

Can you post the contents of the routing table when it doesn't work? It would be under Diagnostics > Routes or from the command line, "netstat -rnW". The interface configuration for the PPPoE interface in the GUI may also help, along with the interface info from "ifconfig pppoe0".

w0w

@jimp:

Can you post the contents of the routing table when it doesn't work? It would be under Diagnostics > Routes or from the command line, "netstat -rnW". The interface configuration for the PPPoE interface in the GUI may also help, along with the interface info from "ifconfig pppoe0".

There is nothing else configured on PPPoE, only username and password. WAN connected directly to ISP ethernet (gigabit). Since I am not using their local LAN, there is no additional interface on WAN, only PPPoE.
Can I PM you screenshot of routing table without messing with removing possible private info? But really it differs only on the first line which is

Destination	Gateway	Flags	Use	Mtu	Netif	Expire
default	212.7.29.236	UGS	124180	1492	pppoe0	
8.8.4.4	212.7.29.236	UGHS	0	1492	pppoe0
.........

If I just completing pfSense wizard the first line appears


Destination	Gateway	Flags	Use	Mtu	Netif	Expire
8.8.4.4	212.7.29.236	UGHS	0	1492	pppoe0

As you can see there is no default route. But even if I select it manually from GUI, then it works untill reboot, also I have already stated that enabling "Use non-local gateway" fixed connection issue completely, even after reboot it is still working.

Here is ifconfig pppoe0

pppoe0: flags=88d1 <up,pointopoint,running,noarp,simplex,multicast>metric 0 mtu 1492
	inet 89.235.x.x --> 212.7.29.253  netmask 0xffffffff
	inet6 fe80::d250:99ff:fec1:71b%pppoe0 prefixlen 64 scopeid 0x7
	nd6 options=21<performnud,auto_linklocal></performnud,auto_linklocal></up,pointopoint,running,noarp,simplex,multicast>

wan.jpg_thumb

w0w

So I did further analysis. I have found that if you have PPPoE configured on old version as WAN, on pre May version, then it takes PPPoE gateway as default gateway, but when you save config.xml there are ZERO entry in <gateways> section, if I change something in my WAN_PPPoE gateway on old April version and save and then change back and save again — then default gateway appears in config.xml and after upgrade to May version I have now internet connection from boot and even after reboot, no need to "Use non-local gateway" option, but if I use new installation and use wizard for creating PPPoE on WAN then I need manually set default route to received via PPPoE ISP gateway appeared in list and then, after next reboot again no Internet copnnection and I need to set "Use non-local gateway" option to make it working constantly.
There are also some images added . The one showing some strange IPv6 information, after manually setting default IPv4 gateway.

This one shows ~~default routes~~ configured gateways after configuring PPPoE from zero on new system.

And at the end there is a comparision of two config files, the one on the left is that is created after manipulating on old version to make appear default gateway and after successful upgrade to May version and on the right pane is that pure May config file AFTER selecting default gateway manually and "Use non-local gateway" option. Both configurations make Internet working. Looks like there is some bug in parsing and writing this section, also there is some bug that prevents setting PPPoE or any dynamic gateway as default on new system.

w0w

I am pretty sure that changes made in this commit https://github.com/pfsense/pfsense/commit/43a9b03deb9db482713dfa1218662bce8b6360ce#diff-1332c372788c9e1a8c6c9bae9ebb55a5
are the root of the problem with default dynamic gateway on PPPoE, in my case. I think it's all around system.inc within deleted lines 647-705, because restoring them fixed the problem with default gateway.
Finally, steps to reproduce the problem. Create your own pfSense firewall, virtually or on real hardware, you will need two Ethernet cards. I have used Auto install on ZFS and after successful install I just used wizard, that automatically started after logging into WEBGUI and I have nothing changed, only selected WAN as PPPoE for IPv4 and that all, after finishing wizard steps you will already have configuration with NO default ROUTE and therefore NO internet connection.

w0w

https://redmine.pfsense.org/issues/8504#change-36668
I thought that it was the similar problem, but also the wrong suspect.

jimp

@w0w said in No internet connection after upgrade from April 20 snapshot to May 19:

https://redmine.pfsense.org/issues/8504#change-36668

That is a completely different gateway issue, unrelated to this one. That was a failed gateway upgrade problem, this is a separate issue having to do with purely dynamic gateways not present in config.xml. If there isn't already a ticket for it, it should have one.

w0w

@jimp, in my case if I follow this closed ticket guidelines, installing the 2.4.3, everything works like a charm, but if I upgrade this working installation to 2.4.4 latest version it just does not have default gateway anymore, but it does not mean your upgrade code fails, yes, and it is different issue, caused by different commit mentioned above. Also, I understand that you have more information about closed issue and you are right in your conclusions, but from my side it looked like similar issue and my conclusions were based just on guideline posted on redmine ticket.
If I understand correctly this removed code that maintained dynamic gateways, making them default without placing them into config.xml, caused some other problems?

jimp

Completely dynamic gateways do not need an entry in config.xml, they never have. They are dynamic and maintained internally. There is nothing in config.xml so there is nothing for the upgrade code to touch.

Again, it's apples and oranges. The symptoms are similar but they are completely different problems.

w0w

@jimp, yes it's absolutely different issue, I see it.
If dynamic gateways don't need entry then why the code maintained them was removed, i mean those system.inc changes?
I am also not sure about redmine tickets regarding this problem, I don't see anything now, but it also can be just because I'm blind

jimp

I don't know, it happened as a part of a larger feature merge, that's what still needs to be investigated. It's possible that it was not intentional, or it wasn't compatible with the new features, etc. That is what remains to be debugged and tested for this specific issue.

jimp

Looking through the code the same actions seem to be taken now but they are split into separate functions. It's still possible there is some difference but I haven't spotted it yet.

I've also tried replicating the problem on several VMs here but I haven't been able to make one end up without a default gateway in the OS routing table. I have a few purely dynamic gateway VMs but they are all DHCP. Since I don't currently have a way to test one that is purely PPPoE, it may be that the issue is specific to only having PPPoE dynamic gateways.

The only thing I have noticed is that despite actually having a default gateway at the OS, these show None in the Default Gateway IPv4 drop-down and nothing in the Default column of the gateway list. There is likely a need to improve the handling of that field. Though I'm still not sure we should even offer a None option for IPv4 at this point. The previous code forced one default at all times.

I need to check into a couple unrelated things now but I'll try to setup a PPPoE only VM if I can and see what happens.

@PiBa may need to have a look as well, he developed the default gateway changes that were merged into 2.4.4.

w0w

@jimp
Thank you! I hope you will find something.

PiBa

Tried a few things but have not yet found a scenario when the default-route gets 'missing'..

Currently testing with 2 pfSense boxes 1 being the pppoe server, the other being the client.. I can change the settings on the server, and the client would loose connection for a few seconds, and then pick it up again, including new gateway-ip and different client-ip..

The gateway pppoe interface is selected as the default on my config in the webgui. Even though it is a 'automatically generated' gateway it does show in the selection boxes, and should be selected to be the default.

The 'non-local gateway' option is should not be related to pppoe connections. It was first created for OVH datacenter usage where they have 'regular' network connections, but do need to route traffic out over the specified interface..

Either way, the gateway selection box being empty would be strange..

Ive also installed a new pfSense and went through the 'wizard' in which case the default-gateway indeed stays as 'none', but other options are shown available in it when clicked, i suppose the wizard needs to be changed to fill the new selection option.. After selecting that WAN_PPPOE manually though the default route gets added properly.

Only thing i could find sofar is that the 'installation wizard' needs a little fix.. When configured, the actual code handling those settings 'seems' to work properly in my tests..

w0w

@piba said in No internet connection after upgrade from April 20 snapshot to May 19:

The ‘non-local gateway’ option

I am also not sure why it's needed in my case, but I think it triggers something else and connection restores. I do think that this can be some other problem, but also related to those changes, because restoring of system.inc fixes everything. I think that we have some scenario like when you are selecting manually this gateway it is then stored in config.xml and then it processed some other way needed this option also, but I am not sure about that. May be the the second one, like adding manually WAN_PPPoE creates one entry in config.xml and checking this ‘non-local gateway’ option just saves my dynamic gateway into config.xml and that's why it works.
I can do some tests tomorrow, I hope.
The one thing that also confuses me it that if I select that WAN_PPPOE manually, it also works without this ‘non-local gateway’ option but only up to re-connection or reboot.

@piba said in No internet connection after upgrade from April 20 snapshot to May 19:

Only thing i could find sofar is that the ‘installation wizard’ needs a little fix…

But this little fix does not solve the problem when you have no default gateway on upgrade, since you have already configured everything and config.xml does not contain dynamic gateway entry.

w0w

Quick test showed that second scenario posted above is true.
No need for this option at all, just pressing 'save' in gateway options do the same thing.

PiBa

Okay seems i can reproduce it now with a reboot while WAN nic is disconnected, then after enabling the wan-nic, it does get a IP, but the default-route isn't set. Then when the PPPOE server side is restarted, the PPPOE client does automatically re-connected and then does set its default-route that time..

Trying to figure out now how where/how the first connection doesn't set the gateway and the second does. Seems to me like both should follow almost the same steps when the connection is established.

PiBa

FYI
what im seeing is this in my systemlog, both times the rc.newwanip is running, but only the second time it tells "Default gateway setting Interface WAN_PPPOE Gateway as default".
I presume you have something similar to the first one in your log.?

Jun 9 15:51:30	php-fpm	321	/rc.start_packages: Restarting/Starting all packages.
Jun 9 15:51:28	check_reload_status		Starting packages
Jun 9 15:51:28	php-fpm	322	/rc.newwanip: pfSense package system has detected an IP change or dynamic WAN reconnection - 10.30.20.163 -> 10.30.20.163 - Restarting packages.
Jun 9 15:51:26	php-fpm	322	/rc.newwanip: Creating rrd update script
Jun 9 15:51:26	php-fpm	322	/rc.newwanip: Resyncing OpenVPN instances for interface WAN.
Jun 9 15:51:23	php-fpm	322	/rc.newwanip: Default gateway setting Interface WAN_PPPOE Gateway as default.
Jun 9 15:51:23	php-fpm	322	/rc.newwanip: rc.newwanip: on (IP address: 10.30.20.163) (interface: WAN[wan]) (real interface: pppoe0).
Jun 9 15:51:23	php-fpm	322	/rc.newwanip: rc.newwanip: Info: starting on pppoe0.
Jun 9 15:51:22	ppp		[wan] IFACE: Rename interface ng0 to pppoe0

Jun 9 15:47:55	php-fpm	73857	/rc.start_packages: Restarting/Starting all packages.
Jun 9 15:47:54	check_reload_status		Starting packages
Jun 9 15:47:54	php-fpm	321	/rc.newwanip: pfSense package system has detected an IP change or dynamic WAN reconnection - 10.30.20.163 -> 10.30.20.163 - Restarting packages.
Jun 9 15:47:52	php-fpm	321	/rc.newwanip: Creating rrd update script
Jun 9 15:47:52	php-fpm	321	/rc.newwanip: Resyncing OpenVPN instances for interface WAN.
Jun 9 15:47:48	php-fpm	321	/rc.newwanip: rc.newwanip: on (IP address: 10.30.20.163) (interface: WAN[wan]) (real interface: pppoe0).
Jun 9 15:47:48	php-fpm	321	/rc.newwanip: rc.newwanip: Info: starting on pppoe0.
Jun 9 15:47:47	ppp		[wan] IFACE: Rename interface ng0 to pppoe0

w0w

@piba
Yep, looks exactly like that. Both. I've just found that on the latest snapshot it works a little bit different way or I just missed it before:
After setting my dynamic gateway manually as default I see that line

/rc.newwanip: Default gateway setting Interface WAN_PPPOE Gateway as default.

then I rebooted firewall and looked at system log again — when PPPoE starts on boot, I don't see this line anymore, system log is exactly like your second one and default route is missing, certainly. If I manually re-connect PPPoE without changing anything else, then I see this line again and everything works like it should be.