I have two machines in a LAN, machine [1] runs Debian Linux and WinXP and machine [2] just runs Debian Linux.
Machine [1] has two harddisks, one with a NTFS partition, the other with an EXT2 partition. While machine [1] boots Debian Linux by default, it mounts the NTFS partition through mount -t ntfs on /mnt/winxp. This works fine: all users on machine [1] can access this partition and see plenty of files and directories.
Machine [2] has one harddisk with just one EXT2 partition. I mounted the root directory (/) of the EXT2 partition of machine [1] through NFS on /mnt/bas-01 on machine [2]. This mount through NFS also works fine except for the mount point /mnt/winxp on machine [1], when I try to access this folder through the mount point /mnt/bas-01 on machine [2] it seems to be empty.

I found just one thread about this kind of problem on an forum elsewhere on the net. The advise to add no_root_squash to /etc/exports didn't help. Also the suggestion that it might be a permission problem is not likely: everyone has read access to this folder and this is confirmed by the fact that everyone on machine [1] can read the folder /mnt/winxp, but as soon as NFS is involved the folder seems to be empty. Does anyone have another suggestion?

Thanks,

Bas

I don't think you can network mount windows partitions with NFS. I think you need to use samba for that.

I'm not trying to mount a windows partition directly. I try to mount a EXT2 partition with a mount point to a windows partition. Or doesn't that make a difference?
I'm not using that windows partition often, so I like to avoid using samba while I'm already using NFS for other purposes.

Bas

NFS don't support Windoze (FAT/FAT32/NTFS) partition... if you mount it to a Ext2 drive, it makes no difference at all, a mount point is just a link to the hard drive... it is just like changing a drive letter on windoze... no matter if your letter is C: or D:, the drive formatting still the same.

that's a pitty, but thanks... i'll try it with samba then...

if i could get my modem to work on linux i would test it, But i cannot see why nfs cannot access fat partitions. If it cannot this is because fat partitions do not use inodes and every part of the linux file system is based on them.

The linux kernel has dynamic inode creation built into the fat layer to fix this problem. This enables the VFS to access it via it's inode_structure.

but one would thing nfsd interfaces with the file system via the VFS, meaning serving via fat should be possible.

But nfsd tends to not follow mounts, meaning if u serve a directory bellow the mount point u will only gain access to the blank directory not the mounting point.

i noticed this on my server as i had a 120Gb mounted on /home/download and mounting /home did not give me access to it, instead i had to mount it as well.

just in general if nfs interfaced with the VFS there is no way mounting a fat volume would be a problem and 2nd there is NO WAY in hell linus would allow nfsd to directly access anything else but the VFS.


So in concluding it is possible to mount a fat partition u just have to explicitly mount the fat mounted point.

hope that helped

thanks a lot! it did work to mount the ntfs mount point directly via nfs. quite stupid of me that i didn't try that... but thanks...

I hate to be a nuisance but there is no way that you can export NTFS partitions under NFS. End of story.

Read /usr/src/linux/Documentation/filesystems/Exporting if you doubt this.

It would be helpful if people did not post unreliable and just plain wrong stuff to this forum. Otherwise we are all wasting our time.

So i was wrong, looks like a file system needs to support mapping of NFS filehandle fragments to dentrys(directory entries). With file systems like fat this would be done in the kernel since these file system are mapped from there fat structure to inodes in the kernel, i automatically assumed support would have been done. I guess one can never assume.

But when you think about it i guess it is a waist of code.

ehh i am 22 i'm bound to make a few mistakes.

Of course it would be VERY useful if you COULD export NTFS partitions under NFS. But I think we are stuck, for the time being, with filesystems like EXT* and the like.

Petition your local kernel developer for improvements to VFS to make all this stuff doable - and transparent. But don't expect quick results - the fs code is not due for reassessment until 2010