Am 11.08.2011 um 15:22 schrieb Dave Love:

> Reuti <[email protected]> writes:
> 
>> Hi,
>> 
>> while playing around with the old (not builtin) startup method of
>> slave tasks involving the SGE delivered version of `rsh` which allows
>> the setting of a port and has the access rights -rws--x--x (in
>> $SGE_ROOT/utilbin/lx24-amd64) I got "permission denied" on a NFSv4
>> mount on a node in the cluster. Falling back to NFSv3 during the mount
>> on the node without changing anything else worked as expected.
>> 
>> First I thought it's because of the SUID, but it turned out to be
>> working again when adding r for group and other.
>> 
>> So the question is: does NFSv4 need the files to be readable and it's
>> just not implemented to serve executable only files? The only thing I
>> found on the net was, that for a local access there is a difference
>> between a pagein and a read of a file. As NFSv3 can't distinguish
>> these different access modes it grants access in any case ignoring the
>> differences.
>> 
>> Can NFSv4 be setup to behave the same way again?
> 
> It doesn't seem to be a general issue.  I can't try with a Linux NFS4
> server, but here's a Red Hat 5 client with a Solaris 10 file server
> (after changing the permissions from the originally-installed ones of
> -rwsr-xr-x):
> 
>  $ ls -l $SGE_ROOT/utilbin/$SGE_ARCH/rsh 
>  -rws--x--x 1 root root 162298 Oct 28  2009 
> /lv1/data/sge/utilbin/lx26-amd64/rsh
>  $ df -T $SGE_ROOT/utilbin/$SGE_ARCH/rsh 
>  Filesystem    Type   1K-blocks      Used Available Use% Mounted on
>  lvfs1:/zpool1/data
>                nfs4   8536599552 2256120832 6280478720  27% /lv1/data
>  $ $SGE_ROOT/utilbin/$SGE_ARCH/rsh 
>  usage: rsh [-nd] [-l login] [-p port] [-t timeout] host command

Thx - so it seems to be on the server side. It's set up in exports as:

/usr/sge                        
@nodes(rw,root_squash,anonuid=63,anongid=58,sync,fsid=1004,subtree_check)

The cluster is in use right now, so I can't play with it again. But I'm still 
curious, whether it's a feature of the Linux kernel (being it intended or a 
bug).
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to