Gianluca Cecchi wrote: > And we come to my initial attempts. > In fact I initially tried to do that, and configured tftpd in xinetd > but I got the same result, so I asked myself if I had to use tftpboot > embedded in Qemu or not. > So, resuming. > - Configured a default bridged nertwork named virbr0 on the host (that > gives ip addresses from 192.168.122.2 - 192.168.122.100), so that > [root@virtfed ~]# ifconfig virbr0 > virbr0 Link encap:Ethernet HWaddr 5E:31:5B:33:F9:5A > inet addr:192.168.122.1 Bcast:192.168.122.255 Mask:255.255.255.0 > UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 > If you can't get this all to work, maybe try using a physical bridge device as opposed to libvirt virtual networking: http://wiki.libvirt.org/page/Networking#Bridged_networking_.28aka_.22shared_physical_device.22.29 People have had issues with virtual networking in the past, so it might be good to eliminate that factor. > - Configured tftpserver on the host. > - From a guest that has its virtual network device on virbr0 I can do > [root@mindy ~]# tftp 192.168.122.1 > tftp> get pxelinux.0 > tftp> quit > [root@mindy ~]# ll pxelinux.0 > -rw-r--r-- 1 root root 15702 Jun 18 12:05 pxelinux.0 > > So tftp works on the host. In its messages I get: > Jun 18 11:59:10 virtfed xinetd[2962]: START: tftp pid=3033 from=192.168.122.102 > > - Trying to create a guest from virt-manager with these steps: > first window I specify guest name and select the option "Network Boot (PXE) > (I can select also Local install media, while the option "Network > install" is greyed out.... is this correct?) > second window I select OS Type=Linux and Version=generic2.6 kernel > third window memory and cpu > fourth window I specify a pre-existing storage (LV on a dedicated VG) > fifth window I do not change anything (I see that are selected virtual > network default virt_type=kvm and os type x86_64, that are ok for me) > I select finish and I get the console window for the guest > The network boot begins and I get this loop > Searching for a Server (DHCP).... > Me: 192.168.122.35, DHCP: 192.168.122.1, TFTP: 192.168.122.1, Gateway > 192.168.122.1 > No filename > <sleep> > <abort> > probing pci nic... > Probing isa nic... > <sleep> > Boot from (N)etwork or (Q)uit? > > probing pci nic... > [rtl8139] - ioaddr 0XC100, irq 11, addr 54:52:00:45:F8:17 100Mbps half-duplex > > And no tftp request in /var/log/messages on the host.... > So I now notice that dhcpd service is provided by dnsmasq on the > host... that probably is started by libvirtd? > because > [root@virtfed ~]# chkconfig --list dnsmasq > dnsmasq 0:off 1:off 2:off 3:off 4:off 5:off 6:off > Correct, libvirtd starts dnsmasq for virtual networks. > And probably the missing part is telling dnsmasq to enable its > embedded tftp-server (or anyway tell it to use an external one...) > So the question is: can I configure this instance of dnsmasq to enable > its tftp server? > I can do it in /etc/dnsmasq.conf, but it seems that dnsmasq is started > without a config file: > > nobody 2428 1 0 13:14 ? 00:00:00 /usr/sbin/dnsmasq > --strict-order --bind-interfaces > --pid-file=/var/run/libvirt/network/default.pid --conf-file= > --listen-address 192.168.122.1 --except-interface lo --dhcp-range > 192.168.122.2,192.168.122.100 > Libvirt doesn't support any tftp options for dnsmasq. > How to glue the several components...? > I've never set up my own pxe server so I can't provide much help. I have though booted many VMs via PXE, with both virtual networking and bridged networking, and it required no extra options. Maybe try and verify your tftp setup is working with a physical machine (if you have the means)? Because there shouldn't need to be much tweaking necessary to get the VM to see it. HTH, Cole _______________________________________________ et-mgmt-tools mailing list et-mgmt-tools@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/et-mgmt-tools
