Have you tried the latest XFS and Samba

John Trostel jtrostel at connex.com
Thu May 3 14:36:26 GMT 2001

The problem I was having was recently fixed by Jeremy in the Samba code.  I'm
not sure but I suspect it was particular to the kind of setup we have here. 
There is no PDC running so the Samba machine was running standalone. Probably a
situation uncommon in bigger NT groups.

Anyway, thanks.  The problem is now fixed.

On 02-May-2001 Juergen Hasch wrote:
> John M Trostel schrieb:
>> You mentioned that Samba & XFS still work together for you. I wonder if
>> it is a strictly winNT problem or not. I downloaded CVS snapshots for
>> the past few weeks and discovered that the last CVS that works with my
>> XFS system was from the CVS no later than April 22. Could you check to
>> see what date your Samba source is from?
> I am using the CVS trees from April 30 (XFS and Samba 2.2). I'll
> try this on a test machine at work. We have a lot of NT machines there.

SGI is pretty close to changing the way acl_get_file and acl_get_fd (the basis
for sys_acl_get_file and sys_acl_get_fd) to return more standard errors.  The
problem being thrashed out now is trying to agree just what the 'standard'
really means (c:

I finally found ENOTSUPP.  It turns out that it is in /usr/include/bit/errno.h,
which gets included in /usr/include/errno.h.  It's just
        #define ENOTSUPP EOPNOTSUPP 

>> I see how your test for acl support using sys_acl_set_{fd/file} would
>> work at the bottom of the function but would still argue for a more
>> universal test at the start. I think SGI will be moving towards changing
>> the acl_get_{fd/file} functions to return EOPNOTSUPP (or maybe ENOSYS)
>> after detecting an acl_set attempt on a non-XFS system. Returning
>> EOPNOTSUPP would be more in line with the posix standard than the
>> current situation. It would then be easier to test as we entered this
>> function.
>> This would also eliminate the need to change the sys_acl_get_file
>> function, as a NULL would be returned if no ACL was present on a file
>> system supporting ACLs & EOPNOTSUP for a file system not supporting
>> ACLs. The Samba posix_acls.c function would need to handle these two
>> different situations.
> No doubt, the best solution is to have the XFS ACLs behave Posix-like.
> SGI seems to move in that direction, so my patches are just meant to 
> get Samba working with ACLs until a clean solution is available.
>> P.S. Is ENOTSUP even defined in Linux? I couldn't find it.
> ENOTSUPP is defined in linux/errno.h
>   ...Juergen

John M. Trostel
Linux OS Engineer
jtrostel at connex.com

More information about the samba-technical mailing list