I am currently using a server which PXE-boots a XEN-based Dom0 image via nfsroot and Unionfs. This works perfectly (with a custom built kernel allowing DHCP autoconfig and nfsroot) on FC4. Upgrading to FC5 works with the (original) 2.6.17-based kernel, but stops due the the known Unionfs problems with 2.6.18.
But the problem is now that I tried to upgrade to FC6 (substituting aufs for unionfs, which works nicely), the NFS rootfs mount shows a very strange behaviour (even before aufs steps in):
While on my 2.6.17 FC5 exporting
/nfsroot/ 172.16.0.0/255.255.254.0(ro,no_root_squash,no_all_squash,asyn c)
yields the following (expected) entry in /proc/mounts of the client
/dev/root / nfs ro,vers=2,rsize=4096,wsize=4096,hard,nolock,proto= udp,timeo=11,retrans=2,addr=172.16.0.1 0 0
the same line with 2.6.18 FC5 (regardless of minor version) gives
/dev/root / nfs ro,vers=2,rsize=4096,wsize=4096,hard,nolock,proto= udp,timeo=11,retrans=2,sec=null,addr=172.16.0.1 0 0
and inhibits access to non-public files (sec=null!!!)
. Exporting via
/nfsroot/ 172.16.0.0/255.255.254.0(ro,root_squash,all_squash,anonuid=0, anongid=0,async)
gives better results, but can't be the solution. Standard NFS mounts work as expected though. Kernel is currently 2.6.18-1.2869.fc6, nfs-utils 1.0.10-5.fc6.
Can anyone offer help or an explanation of what's happening