Ask Your Question

Fedora25 KVM hypervisor macvtap i/f with FedoraCore 6 guest not working

asked 2017-02-16 04:38:42 -0600

cwicz gravatar image

Hi all,

I have Xen FedoraCore6 Xen hypervisor with a single fully virtualized FedoraCore6 guest. FedoraCore6 hypervisor has 3 bridges:

homenet         8000.d6557dc33b2f       no              tap0
internet                8000.e283ec832f8e       no              vif5.1
newinternet             8000.02e0868f5ae2       no              vif5.2

The FedoraCore6 guest, hosts my http/smtp services and routes all traffic between homenet and internet. Xen host and guest have Kernel 2.6.18-1.2798. This has worked perfectly for 10 years with hardly any changes, but due to recent blown PSU need to upgrade aging hardware..

When I launch a copy of FedoraCore6 guest on Fedora25 KVM hypervisor on a new laptop, replacing the bridges with macvtap bridges, internet connection (from "homenet") to youtube or speedtest, initially connect but after 30 seconds of burst activity everything grinds to a halt; speedtest bombs out with socket errors and all routed internet connectivity is lost.

On the FedoraCore6 guest I find that I can no longer ping the internet router via I/F eth1/eth2 (enp0s20u1/enp0s20u2) macvtap bridges, the FedoraCore6 guest becomes isolated in someway. I can ping internet router from Fedora25 KVM hypervisor using enp0s20u1/enp0s20u1 interfaces but can nolonger ping internet router from FedoraCore6 guest using eth1/eth2 macvtap bridged interfaces.

I have tried the following on Fedora25 KVM hypervisor but always get same results:

ifconfig enp0s20u2 promisc ### ("new internet") / NM_CONTROLLED=no 
                           ### in /etc/sysconfig/network-scripts
ifconfig enp0s20u1 promisc ### ("internet)") / NM_CONTROLLED=no 
                           ### in /etc/sysconfig/network-scripts
ifconfig eno1 promisc      ### ("homenet") / NM_CONTROLLED=no
                           ### in /etc/sysconfig/network-scripts
ifconfig eno1
ifconfig enp0s20u1
ifconfig enp0s20u2
ifconfig enp0s20u2 up
ifconfig enp0s20u1 up
ifconfig eno1 up
firewall-cmd --set-default-zone=trusted
firewall-cmd --permanent --direct --passthrough ipv4 -I FORWARD -m physdev
 --physdev-is-bridged -j ACCEPT
firewall-cmd --reload
iptables -F

Also tried disabling SELinux, firewalld on Fedora25 KVM hypervisor.

No errors in /var/log/messages on Fedora25 KVM hypervisor or FedoarCore6 guest.

Fedora25 4.8.6-300.fc25.x86_64 KVM hypervisor + FedoraCore6 2.6.18-1.2798 guest (rtl8139 NIC's).

Anyone have any ideas please?


Best Regards - Colin

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2017-02-16 16:49:36 -0600

cwicz gravatar image

Hi all,

Whilst writing the initial question, I used google to verify "rtl8193" model and by bizarre coincidence, a google search using: "kvm rtl8193", returned the following top hit:

This documentation suggests using e1000 driver not rtl8139 for 10Gb/s networks.

My network is 1Gb/s, but having changed my FedoaCore6 guest /etc/modprobe.conf aliases from rtl8139 to e1000 and changing Fedora25 KVM hypervisor guest config (using virsh edit) to also use e1000, everything now works normally.

So there appears to be an issue with rtl8139 driver, it's not just a performance or 10Gb/s issue as documented as connections seem to silently drop/disconnect.

Hope this helps someone, wasted many hours since last Saturday trying to resolve..

Best Regards - Colin

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2017-02-16 04:38:42 -0600

Seen: 230 times

Last updated: Feb 16 '17