On Sun, 2 Jul 2017 21:12:17 +0200
Milan Obuch <freebsd-curr...@dino.sk> wrote:

> On Sun, 2 Jul 2017 20:13:49 +0200
> "Hartmann, O." <o.hartm...@walstatt.org> wrote:
> 
> > On Sun, 2 Jul 2017 14:39:34 +0200
> > Milan Obuch <freebsd-curr...@dino.sk> wrote:  
> 
> [ snip ]
> 
> > > > To not use a routing daemon due to the small size of my network, I
> > > > desided to use static routes, in rc.conf I placed the following
> > > > variables:
> > > > 
> > > > static_routes="igb1.2 igb1.10"
> > > > route_igb1_2="-net 192.168.2.0/24 -interface igb1.2"
> > > > route_igb1_10="-net 192.168.10.0/24 -interface igb1.10"
> > > > 
> > > > igb1 is assigned to IP/NET 192.168.0.1/24
> > > >   
> 
> Just to be exact, could you show us ifconfig lines from rc.conf as well?
> It is common to have something like
> 
> cloned_interfaces="igb1.2 igb1.10"
> ifconfig_igb1_2="192.168.2.1/24"
> ifconfig_igb1_10="192.168.10.1/24"
> 
> and no static routes as you showed, because address assigned to
> interface means automatically line in route table, however, they should
> look identical to those shown in your first mail.
> 
> > > > netstat -Warn gives me (as dummy, since I have no direct access to
> > > > the box via serial console from the system I write this mail):
> > > > 
> > > > Internet:
> > > > Destination      Gateway         Flags       Use    Mtu      Netif
> > > > 127.0.0.1        link#3          UH       334564  16384        lo0
> > > > 192.168.0.0/24   link#4          U         23452   1500
> > > > igb1 192.168.0.1      link#4          UHS       29734
> > > > 16384        lo0 192.168.2.0/24   link#5          U
> > > > 271   1500 igb1.2 192.168.2.1      link#5          UHS           0
> > > > 16384        lo0     
> > > 
> > > I think you did not include network 192.168.10.0/24 on igb1.10...    
> > 
> > I skipped that, it is quite the same according to the settings of the
> > others and unused for now. So it doesn't matter. But you're right.
> >  
> 
> This was just for tha sake of completteness, nothing else.
> 
> [ sysctl stuff snipped - not relevant, I think ]
> 
> > > > From the routing device itself, it is possible to ssh into a VoIP
> > > > client attached to the switch to which igb1.2 trunks the net.
> > > > Pinging is also possible.
> > > > 
> > > > Attached to igb1 is the 192.168.0.1/24 network with a bunch of
> > > > hosts. From any host within this network it is possible to ping
> > > > the 192.168.2.0/24 network and its hosts within, but no SSH, not
> > > > web (80, 443). 
> > > >      
> > > 
> > > Weird - if icmp (ping) works and tcp (web, ssh) not, something is
> > > filtering traffic. But with net.inet.ip.forwarding=0, even pinging
> > > host should not work. Try tcpdump to see what's going on.     
> > 
> > net.inet.ip.forwarding works as expected. See above, I confused the
> > OID.
> >   
> 
> [ snip ]
> 
> > > From network architecture view, there is no difference - vlan is
> > > network interface just like physical ethernet. Basically everything
> > > is the same (sometimes there is issue with mtu, but this hardware
> > > dependent).    
> > 
> > Yes, so I thought, but as you stated, something is filtering and I
> > have no clue what.
> >   
> 
> Then I just recommend tcpdump - I would use 'tcpdump -nepi igb1.2 host
> 192.168.0.x and host 192.168.2.y' and 'tcpdump -nepi igb1 host
> 192.168.0.x and host 192.168.2.y' in two session and compare outputs
> when pinging from 192.168.0.x to 192.168.2.y and when trying to ssh
> from the former to the later. Also there is a question then what these
> two devices are, what OS are they running, their network
> configuration... then we can analyse the problem better.
> 
> Regards,
> Milan
[...]

Well, some news from a "lost" night at the HomeOfficeFrontier.

I followed the advices given by you (Milan and Freddie), except the tcpdump
sessions, because I also had some trouble with the ISP's connection.

But: Having setup the router's interface to igb1.10 (vlan 10) revealed some
serious problems with the setup of the switch I use in the HomeOffice. We use
mostly Cisco switches. It is easy to assign ports to a certain VLAN and leave
them "untagged", but the uplink port (Cisco calls this port trunk port or
etherport) has, of course, "tagged" etherframes.

The switch is a Netgear GS110TP, the uplinkport is g9 (SFP copper). This
tagged port is attached via CAT 6 cable to the igb1 of the router. The router,
the FreeBSD 12-CURRENT box in question here, has VLANs 2, 10, 66 and 100
assigned to this port, but I use only 2 and 10 at the moment. vlan 2 is, as
explained above, the VoIP network, its switchport is g8 which "must" be tagged
to reach the Grandstream VoIP phone, which has 802.1q tag 2. 

So, as Freddie Cash suggested, I assigned my native LAN (192.168.0.1/24) to
igb1.10 and assigned the uplink port of the switch also to be member of vlan
10 "tagged" and put also the other ports (3 ports) with hosts attached to the
net 192.168.0.1/24 into the group of VLAN 10. 

The moment I put a port with a host in vlan group 10 out of the default group
(vlan 1), the host is not reachable anymore. This confuses me and I think the
problem is more related to the weird Netgear stuff than FreeBSD. At least, the
router should send ICMP packages tagged with vlan 10 via the trunk port to its
vlang 10-gruoped (untagged) ports, on which a freeBSD host is listening with an
interface not assigned to a vlan. I have not checked what happens if the one
host of this group has its NIC put into the vlan 10. There is probably a
different handling of Ingress and Egress etherframes on that Netgear GS110TP
switch which I have not thought of.

Anyway, thank you very much for helping.

Kind regards,
Oliver 
_______________________________________________
freebsd-current@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscr...@freebsd.org"

Reply via email to