beta2

[mk01]

(15th Oct, 2013 06:35 AM)Dizzy Wrote:  setting GSSD and IDMAPD as yes by default harms NFSv3 compatibility?

Big kudos for 1.2, it's a huge improvement over 1.1 which quite literally made my stomach ache

thanks for reporting back.

I'm just fixing for combo nfs and nfsvers=4 in mount options. will you be ok to test on request by removing "yes" from nfs-common, leaving nfs as fstype in fstab but adding nfsvers=4 to mount options to fstab?

GSSD and IDMAPD will not affect NFS2/3 in any way. on one side it is still NFS, but working own way. you don't need even rpcbind for it, no mountd daemon to provide locking etc (so NFS roots are benefiting from this). they are designed to coexist without issues and also work independently if needed. just /etc/exports file definition can be tricky a bit as there is no direct flag in to specify which is which so if you don't specify in client neither one can run into troubles like yours was.

so it is possible to provide the same directory for NFS2/3 as export with different paths as for NFS4. this are two lines from exports file which will allow this (the folder under /expornfs are just bound from original locations, not physically mounted there):

Code:

/exportnfs        192.168.1.0/24(rw,insecure,no_subtree_check,crossmnt,fsid=0)
/exportnfs/xbmc        192.168.1.0/24(rw,async,no_subtree_check,all_squash,insecure,anonuid=501,anongid=4,sec=sys,n​o_wdelay)
/exportnfs/media    192.168.1.0/24(rw,async,no_subtree_check,all_squash,insecure,anonuid=501,anongid=4,sec=sys,n​o_wdelay)

this will allow NFS4 browsing and filesystem crossing and server browsing with accessing just the server like media:/ or my_ip:/ . this is NFS4 root where you will see all the export provided by NFS server. this wasn't possible with NFS2/3. you had to mount specifically /media or /xbmc. so you had to know the exports before.

for NFS2/3 previous rules apply so export is to be called media:/exportnfs/media media:/exportnfs/xbmc. so no browsing there.

with the fsid=0 you specifies so called NFS4 root. this is the dir where NFS4 will start if don't specify what you mounts and will allow browsing, referencing, even cross mounting.

and by using media:/xbmc or media:/exportnfs/xbmc as source definition for mount command the nfs client/server negotiates the right protocol - NFS4 for the names which are at root relative to fsid=0 (so like :/media) and NFS2/3 for the full names like /exportnfs/xbmc.

unfortunately this theory around is not clearly described in empty /etc/exports file nor on most wikis/howtos I have seen. but can provide significant saving in complexity of autofs definitions, disk less setups and others. as when you want to add additional export, you just creates new dir under fsid=0, binds the location to it and you are done. you don't to change clients.

so your setup would change like:
1) only one mount to take care of in /etc/fstab in RPI (10.0.0.1://home_export)
2) only one record in /etc/exports ((in case of pure NFS4)):
"/home_export 192.168.1.0/24(rw,no_subtree_check,crossmnt,fsid=0)"
3) then to add / remove Movies / Series / Music would mean
- creating dir /home_export/Movies
- bonding original /home/_Movies to it (mount -o bind /home/_Movies /home_export/_Movies
4) that's IT ! no change on chlient, no change on server in /etc/exports, no need to run "exportfs -a". I know, for three fs this is no difference. But for large systems previous NFS2/3 definitions / workflows have been nightmare to administer.