On Wed, 28 Jul 2004 21:30:09 +0530, mitul@mitul.com mitul@mitul.com wrote:
Hi Raseel,
Usually this is a bad idea.
You can have the nfsroot put up in the kernel command line.
append root=/dev/nfs nfsroot=a.b.c.d:/path/to/nfs/root/on/the/nfs_server/
Look at this page : http://www.tldp.org/HOWTO/Diskless-root-NFS-other-HOWTO-4.html
Best Regards, Mitul Limbani Quoting Raseel Bhagat raseelbhagat@rediffmail.com:
=0A =0AHello Luggers, =0A =0ADoes anyone know how to do = a NFS mount WITHOUT using tftp or dhcp or bootp? =0A =0AI have a kern= el compiled on a client machine having NFS support. I have a Embedded= root filesystem ready on the Nfs server and I have already exported its co= ntents. =0A =0Aportmap, nfsd and mfslock are running on the Nfs serve= r. =0A =0ANow, if I mistake not, by passing appropriate arguments to = the bootloader (i.e.Grub in my case), we can make the kernel from the clien= t machine boot and mount the remote root filesystem(residing on the server)= =0A =0A =0AIs this actually possible? Can anyone provide me the l= inks?
Why is this a bad idea ?? I think it is a fairly good idea if you are using embedded systems or thin-clients. Other than the performance hits due to NFS over udp. (you can configure nfs to work over tcp as well), there should not be any other disadvantage.