I have a problem in communication between the host and the nested VM.This is my layout.
I have installed KVM on host machine having a single nic interface (public ip) running Ubuntu.On top of this,I have VM running Ubuntu.I have installed KVM in this VM too.I then have a VM inside this running a web server. I am able to ping the host from this web server VM and ssh into it.But from host to VM ,ping is being unsuccessful.

The VM (named L1hyp) on host was created using libvirt-manager and has IP of 192.168.122.8. The vswitch interface created at host is in default config (NAT-ed). Its IP is 192.168.122.1.
Now this VM is also having a vswitch interface which is in default config (NAT-ed).Its IP is 192.168.100.1. The Web server VM is created on top of this L1hyp VM, is having an IP of 192.168.100.186. The Webserver VM uses 192.168.100.1 as its default gw. The L1hyp uses 192.168.122.1 as its default gw.

From Host:
ping 192.168.122.8 - SUCCEEDS
ping 192.168.122.1 - SUCCEEDS
ping 192.168.100.1 - SUCCEEDS
ping 192.168.100.186 - FAILS
Comes up with Destination Host Unreachable From 192.168.122.1
But there is route to 192.168.100.0/24 subnet from host.Ping to 192.168.100.1 succeeds.

From Webserver VM:
ping 192.168.100.1 - SUCCEEDS
ping 192.168.122.1 - SUCCEEDS
SSH from web server VM to host succeeds.

Can anyone help me out what needs to be modified to have two way communication between the host and Webserver VM at the earliest?
I am pondering over this problem for over a week now..

Running a VM inside another VM (nested) isn't supported by most (if not all) VM's. I personally don't see the point in doing so....

That said: I have seen people do this, but a lot of trail and error was needed to get it to work.

Do a internet search (example: linux vm inside vm) if you still want to nest VM's. Maybe you'll get lucky and someone has succeeded in setting up your scenario.

can you ping the WebserverVM from your L1hyp?
if you would sniff for IP packets on all your machines, can you follow the ICMP packet path all the way through, e.g. see it coming in into WebserverVM, travelling L1hyp, leaving host and all the way back?

Usually your scenario should just work. I'm using similar setups (vm in vm) wuite often while setting up test servers in virtual environments, which is a clone of a production server running on bare metal and hosting a few VMs.
This can certainly done by KVM (I'm using RHEL/CentOS5.9/6.4)

Hello,
When I did a tcpdump for a ping requests to Webserver VM, on eth0 of L1Hyp (192.168.122.8 connected to 192.168.122.1) ,I am able to see the ping request from the host. I did a tcpdump in virbr0 of L1Hyp (192.168.100.1) to which the Web Server VM is connected. I don't see any requests being forwarded.
Can you please let me know the iptables rules that u have on both host and L1hyp?

looks like your virtual bridge within L2hyp isn't connected to L1hyp's eth0 both ways.

Check if you have IP forwarding enabled.

it could be down to IP tables, but doesn't need to.

My IPtables settings wouldn't help you much since I'm not using the virtbr0 nat interfaces which come with KVM by default but some manual bridging.

The ip forwarding is enabled in the L1hyp.
When I did a tcpdump on the eth interface of L1hyp, I see arp req of "who-has 192.168.100.186 tell 192.168.122.1"
On the virbr0 of L1hyp,I don't see anything.