« Back to Technical Questions

RE: VLAN on eth1 and RAW sockets

Combination View Flat View Tree View
It seems there is no native ability to host a proper VLAN interface off of eth1 on the blade due to the fact that the 801q lkm is not available out of box. Moreover, configuring a vlan interface via the SM CLI works fine (ie. eth1.3) and the interface is active in the Guest OS. However, when it comes to retreiving data off of the interface via a mechanism such as raw sockets, ioctl's such as poll() stay quiet forever, irrepecitve of data being sent there. Switching to a normal virtual interface works fine. Can you confirm this?
 
-J

VLAN is built into the AXP kernel.  That¿s why no lkm is required. VLAN should work for eth0 as well as eth1.

A few questions about the scenario.
1)    Is eth1 configured as part of the VLAN, if so, can the vlan interface be pinged?

2)    What is replicating the traffic to eth1¿s VLAN?  Just enable raw socket does not automatically inform the upstream switch to forward packet to the interface.

at the SM level eth1 is registered as a vlan interface
 
interface eth1.3
ip address X.X.X.X 255.255.255.0
end
 
app-service X
bind interface eth1.3
end
 
In Guest OS, eth1.3 is up and available with X.X.X.x. Sending packets to X.X.X.X is verified using tcpdump to X.X.X.X in the Guest OS
 
yes ping works fine.
 
poll() sits idle no matter the amounf of data pumped to X.X.X.X.
 
Switching everything above to eth1:3 (no vlan), it all works as expected. I originally thought that the absense of 8021q lkm prevented vlan on eth1 and only allowed it via eth0 via some special IOS hidden bridge...but you say its built in.
 
perhaps it s a route issue on inbound where packets are going to the parent interface (eth1)
 
There are a few other virtual interfaces on eth1. Everything shares the same C subnet, perhaps that is the issue related to routing
 
2)   
What is replicating the traffic to eth1¿s VLAN?  Just enable raw socket
does not automatically inform the upstream switch to forward packet to
the interface.
Not sure what you mean by replicating...Systems are directly routing UDP traffic, in this case RTP to X.X.X.X. Guest OS confirms receiving it via aforementioned.
 
J

I am going to triple check all the networking next week to be sure eveyrthing is being tagged properly. It could be a non-issue...
thanks
J

We've looked into the matter and don't see an issue with the VLAN capturing packets which we've verified using tcpdump.  Our team has provided a few references below:
 
-- Not knowing what the developer is doing please have him review:
http://www.tcpdump.org/pcap.htm
http://yuba.stanford.edu/~casado/pcap/section1.html
http://www.binarytides.com/blog/packet-sniffer-code-in-c-using-linux-sockets-bsd/
 
 

I found the issue. There was anetworking caveat.