Hi, I was hoping someone on this forums could help me out with a problem I was facing with Xen. When I compiled the kernel for Xen-3.1.3 using "make world"(I know this is an old version but I need to use it for some compatibility issues), the kernel was missing the Xen netloop module. There was no /kernel/drivers/xen/netback/netloop.ko in the newly compiled kernel. I think we would need that module to enable the bridging between the VM and the physical machine interfaces so that we can migrate the VM. Does anyone know how to solve this problem? or a workaround? Thanks Virajith _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
--- On Wed, 1/5/11, Virajith Jalaparti <virajith.j@gmail.com> wrote:> From: Virajith Jalaparti <virajith.j@gmail.com> > Subject: [Xen-users] Xen Netloop module missing > To: xen-users@lists.xensource.com > Date: Wednesday, January 5, 2011, 3:27 PM > Hi, > > I was hoping someone on this forums could help me out with > a problem I was facing with Xen. > When I compiled the kernel for Xen-3.1.3 using "make > world"(I know this is an old version but I need to use it > for some compatibility issues), the kernel was missing the > Xen netloop module. There was no > /kernel/drivers/xen/netback/netloop.ko in the newly compiled > kernel. I think we would need that module to enable the > bridging between the VM and the physical machine interfaces > so that we can migrate the VM. Does anyone know how to solve > this problem? or a workaround?grep -i xen /boot/config-2.6.18-xxyy | grep -i loop grep your kernel config for the following (and prove its compiled in and not a module) CONFIG_XEN_NETDEV_LOOPBACK=y -- Mark _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi, The kernel config I am using does have the line "CONFIG_XEN_NETDEV_LOOPBACK=y". I am not sure I understand what you meant by "prove it is compiled in and not a module". Are you saying that I am to verify this is indeed the config that is being used? Is so, could you let me know how I can verify that? (Attached is the config file that I was using). Thanks! On 1/5/2011 5:50 PM, Mark Pryor wrote:> > --- On Wed, 1/5/11, Virajith Jalaparti<virajith.j@gmail.com> wrote: > >> From: Virajith Jalaparti<virajith.j@gmail.com> >> Subject: [Xen-users] Xen Netloop module missing >> To: xen-users@lists.xensource.com >> Date: Wednesday, January 5, 2011, 3:27 PM >> Hi, >> >> I was hoping someone on this forums could help me out with >> a problem I was facing with Xen. >> When I compiled the kernel for Xen-3.1.3 using "make >> world"(I know this is an old version but I need to use it >> for some compatibility issues), the kernel was missing the >> Xen netloop module. There was no >> /kernel/drivers/xen/netback/netloop.ko in the newly compiled >> kernel. I think we would need that module to enable the >> bridging between the VM and the physical machine interfaces >> so that we can migrate the VM. Does anyone know how to solve >> this problem? or a workaround? > grep -i xen /boot/config-2.6.18-xxyy | grep -i loop > > grep your kernel config for the following (and prove its compiled in and not a module) > CONFIG_XEN_NETDEV_LOOPBACK=y >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Thu, Jan 6, 2011 at 10:01 AM, Virajith Jalaparti <virajith.j@gmail.com> wrote:> Hi, > > The kernel config I am using does have the line > "CONFIG_XEN_NETDEV_LOOPBACK=y". I am not sure I understand what you meant by > "prove it is compiled in and not a module".It means if you need xen netloop support, you need to have it compiled either as module (CONFIG_XEN_NETDEV_LOOPBACK=y) or builtin (CONFIG_XEN_NETDEV_LOOPBACK=y). If it''s neither, then you can''t use it. You need to recompile the kernel. Also, which Xen version are you using? Are you using the default network-bridge or create your own bridges? netloop is primarily needed for older versions of Xen where the default bridged setup is : - create virtual network device pairs for each dom0 interface (vif0.x and vethx) using netloop module - rename dom0 interfaces (eth0 -> peth0, veth0 -> eth0) - create a bridge (xenbr0) with peth0 and vif0.0 as its members. Newer bridge setup use "eth0" as bridge name, and doesn''t need netloop anymore. It has nothing to do whatsoever with VM migration. In short, if in your current setup your domU is able to access outside network, most likely you don''t need netloop support. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> It means if you need xen netloop support, you need to have it compiled > either as module (CONFIG_XEN_NETDEV_LOOPBACK=y) or builtin > (CONFIG_XEN_NETDEV_LOOPBACK=y). If it''s neither, then you can''t use > it.Do both the cases (compiled as a module or builtin) have the same configuration? (CONFIG_XEN_NETDEV_LOOPBACK=y)> You need to recompile the kernel.Is this as easy as just doing a "make world"? Would that ensure that the kernel is re-compiled? The thing is when I do a "make world" the /boot/config-* file is re-created. How can I be sure that the new kernel is compiled with this flag enabled i.e. with the netloop module builtin?> Also, which Xen version are you using? Are you using the default > network-bridge or create your own bridges? > netloop is primarily needed for older versions of Xen where the > default bridged setup is : > - create virtual network device pairs for each dom0 interface (vif0.x > and vethx) using netloop module > - rename dom0 interfaces (eth0 -> peth0, veth0 -> eth0) > - create a bridge (xenbr0) with peth0 and vif0.0 as its members. > > Newer bridge setup use "eth0" as bridge name, and doesn''t need netloop anymore. > > It has nothing to do whatsoever with VM migration. > > In short, if in your current setup your domU is able to access outside > network, most likely you don''t need netloop support. >I am using Xen-3.1.3 and I was creating my own bridges. In particular, I was using the following set of commands to set up the bridges in the original version of xen I had installed (from yum repos in CentOS). vconfig add peth2 513 brctl addbr breth2.513 brctl setfd breth2.513 0 ip link set dev breth2.513 arp off multicast off up brctl addif breth2.513 peth2.513 ip link set dev peth2.513 up ip link set dev vif0.4 arp off multicast off brctl addif breth2.513 vif0.4 ip link set dev vif0.4 up ip link set dev veth4 name eth2.513 mtu 1400 ip link set dev eth2.513 address `cat /sys/class/net/eth2/address` ifconfig eth2.513 $ip netmask 255.255.255.0 up _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Thu, Jan 6, 2011 at 10:48 AM, Virajith Jalaparti <virajith.j@gmail.com> wrote:> >> It means if you need xen netloop support, you need to have it compiled >> either as module (CONFIG_XEN_NETDEV_LOOPBACK=y) or builtin >> (CONFIG_XEN_NETDEV_LOOPBACK=y). If it''s neither, then you can''t use >> it. > > Do both the cases (compiled as a module or builtin) have the same > configuration? (CONFIG_XEN_NETDEV_LOOPBACK=y)Typo on my part. When built as module, CONFIG_XEN_NETDEV_LOOPBACK=m> >> You need to recompile the kernel. > > Is this as easy as just doing a "make world"?Juding from your questions, the answer would be "no".> Would that ensure that the > kernel is re-compiled? The thing is when I do a "make world" the > /boot/config-* file is re-created. How can I be sure that the new kernel is > compiled with this flag enabled i.e. with the netloop module builtin?Look at documentation (README, etc) found in xen source tarball. I think something like make linux-2.6-xen0-config CONFIGMODE=menuconfig is necessary for 2.6.18 kernel. Note that there are other kernels available for dom0: http://wiki.xensource.com/xenwiki/XenDom0Kernels and with newer (2.6.3x) kernels the process is similar with the usual kernel compile process: make menuconfig, make, and so on. You just need to enable xen-related options during menuconfig process.> >> Also, which Xen version are you using? Are you using the default >> network-bridge or create your own bridges? >> netloop is primarily needed for older versions of Xen where the >> default bridged setup is : >> - create virtual network device pairs for each dom0 interface (vif0.x >> and vethx) using netloop module >> - rename dom0 interfaces (eth0 -> peth0, veth0 -> eth0) >> - create a bridge (xenbr0) with peth0 and vif0.0 as its members. >> >> Newer bridge setup use "eth0" as bridge name, and doesn''t need netloop >> anymore. >> >> It has nothing to do whatsoever with VM migration. >> >> In short, if in your current setup your domU is able to access outside >> network, most likely you don''t need netloop support. >> > I am using Xen-3.1.3 and I was creating my own bridges. In particular, I was > using the following set of commands to set up the bridges in the original > version of xen I had installed (from yum repos in CentOS).If you use Centos, why would you need Xen 3.1.3, and why would you need to compile your own kernel? RHEL/Centos 5.5 comes with Xen 3.1.2+, and has a functional kernel-xen by default.> vconfig add peth2 513 > brctl addbr breth2.513 > brctl setfd breth2.513 0 > ip link set dev breth2.513 arp off multicast off up > brctl addif breth2.513 peth2.513 > ip link set dev peth2.513 up > ip link set dev vif0.4 arp off multicast off > brctl addif breth2.513 vif0.4 > ip link set dev vif0.4 up > ip link set dev veth4 name eth2.513 mtu 1400 > ip link set dev eth2.513 address `cat /sys/class/net/eth2/address` > ifconfig eth2.513 $ip netmask 255.255.255.0 upYou''re mixing xen''s default bridge script (which creates peth and vif0.x, and need netloop support) with your own setup. Don''t do that. Just do everything using OS config files. For example, in my setup (I use eth4 and vlan129. Adjust interface names, vlans, MAC, and IP with your setup) - comment-out network-script line on /etc/xen/xend-config.sxp - some config files: # cat /etc/sysconfig/network-scripts/ifcfg-eth4 DEVICE=eth4 BOOTPROTO=static HWADDR=00:XX:XX:XX:XX:XX ONBOOT=yes # cat /etc/sysconfig/network-scripts/ifcfg-eth4.129 DEVICE=eth4.129 VLAN=yes BOOTPROTO=none ONBOOT=yes BRIDGE=br129 # cat /etc/sysconfig/network-scripts/ifcfg-br129 DEVICE=br129 TYPE=Bridge BOOTPROTO=static IPADDR=192.168.129.xxx GATEWAY=192.168.129.1 NETMASK=255.255.255.0 ONBOOT=yes - assign domU to use br129 (or whatever bridge you use), with something like this on your domU config file vif = [ ''mac=00:16:3E:9A:89:3D, bridge=br129, vifname=myDomU-eth0'' ] Note that in this kind of setup you don''t need netloop support. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> Look at documentation (README, etc) found in xen source tarball. I > think something like > > make linux-2.6-xen0-config CONFIGMODE=menuconfig > > is necessary for 2.6.18 kernel. > > Note that there are other kernels available for dom0: > http://wiki.xensource.com/xenwiki/XenDom0Kernels > and with newer (2.6.3x) kernels the process is similar with the usual > kernel compile process: make menuconfig, make, and so on. You just > need to enable xen-related options during menuconfig process. >I did this and the config file has netloop built in (CONFIG_XEN_NETDEV_LOOPBACK=y) but I cannot find the netloop.ko file in /lib/modules/2.6.18.8/kernel/xen/drivers/. Is there anyway otherway I can verify that this option is indeed being taken into account?>>> Also, which Xen version are you using? Are you using the default >>> network-bridge or create your own bridges? >>> netloop is primarily needed for older versions of Xen where the >>> default bridged setup is : >>> - create virtual network device pairs for each dom0 interface (vif0.x >>> and vethx) using netloop module >>> - rename dom0 interfaces (eth0 -> peth0, veth0 -> eth0) >>> - create a bridge (xenbr0) with peth0 and vif0.0 as its members. >>> >>> Newer bridge setup use "eth0" as bridge name, and doesn''t need netloop >>> anymore. >>> >>> It has nothing to do whatsoever with VM migration. >>> >>> In short, if in your current setup your domU is able to access outside >>> network, most likely you don''t need netloop support. >>> >> I am using Xen-3.1.3 and I was creating my own bridges. In particular, I was >> using the following set of commands to set up the bridges in the original >> version of xen I had installed (from yum repos in CentOS). > If you use Centos, why would you need Xen 3.1.3, and why would you > need to compile your own kernel? > RHEL/Centos 5.5 comes with Xen 3.1.2+, and has a functional kernel-xen > by default. >I wanted to install Xen from source code because of changes that I need to do to it. The repo version of Xen for CentOS works quite well. I do have Xen already installed from repos and trying to install another one from source code.>> vconfig add peth2 513 >> brctl addbr breth2.513 >> brctl setfd breth2.513 0 >> ip link set dev breth2.513 arp off multicast off up >> brctl addif breth2.513 peth2.513 >> ip link set dev peth2.513 up >> ip link set dev vif0.4 arp off multicast off >> brctl addif breth2.513 vif0.4 >> ip link set dev vif0.4 up >> ip link set dev veth4 name eth2.513 mtu 1400 >> ip link set dev eth2.513 address `cat /sys/class/net/eth2/address` >> ifconfig eth2.513 $ip netmask 255.255.255.0 up > You''re mixing xen''s default bridge script (which creates peth and > vif0.x, and need netloop support) with your own setup. Don''t do that. > > Just do everything using OS config files. For example, in my setup (I > use eth4 and vlan129. Adjust interface names, vlans, MAC, and IP with > your setup) > > - comment-out network-script line on /etc/xen/xend-config.sxp > - some config files: > # cat /etc/sysconfig/network-scripts/ifcfg-eth4 > DEVICE=eth4 > BOOTPROTO=static > HWADDR=00:XX:XX:XX:XX:XX > ONBOOT=yes > > # cat /etc/sysconfig/network-scripts/ifcfg-eth4.129 > DEVICE=eth4.129 > VLAN=yes > BOOTPROTO=none > ONBOOT=yes > BRIDGE=br129 > > # cat /etc/sysconfig/network-scripts/ifcfg-br129 > DEVICE=br129 > TYPE=Bridge > BOOTPROTO=static > IPADDR=192.168.129.xxx > GATEWAY=192.168.129.1 > NETMASK=255.255.255.0 > ONBOOT=yes > > - assign domU to use br129 (or whatever bridge you use), with > something like this on your domU config file > > vif = [ ''mac=00:16:3E:9A:89:3D, bridge=br129, vifname=myDomU-eth0'' ] > > Note that in this kind of setup you don''t need netloop support. >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Fri, Jan 7, 2011 at 2:57 AM, Virajith Jalaparti <virajith.j@gmail.com>wrote:> > Look at documentation (README, etc) found in xen source tarball. I >> think something like >> >> make linux-2.6-xen0-config CONFIGMODE=menuconfig >> >> is necessary for 2.6.18 kernel. >> >> Note that there are other kernels available for dom0: >> http://wiki.xensource.com/xenwiki/XenDom0Kernels >> and with newer (2.6.3x) kernels the process is similar with the usual >> kernel compile process: make menuconfig, make, and so on. You just >> need to enable xen-related options during menuconfig process. >> >> > I did this and the config file has netloop built in > (CONFIG_XEN_NETDEV_LOOPBACK=y) but I cannot find the netloop.ko file in > /lib/modules/2.6.18.8/kernel/xen/drivers/. > Is there anyway otherway I can verify that this option is indeed being > taken into account? >If you compile with CONFIG_XEN_NETDEV_LOOPBACK=y then you shouldn''t find netloop.ko file at all as it is not build as a module. Thanks. Kindest regards, Giam Teck Choon _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Possibly Parallel Threads
- Problems with installing Xen 3.1.3 from source code.
- Compiling Xen from source on a machine without Internet access
- Xen migration fails
- No peth interface for xenbr4 and xenbr5
- Bug#388486: xen-utils-common: upgrade new 2.6.17-2-xen breaks create domU until modprobe netloop nloopbacks=n