smbfs: file link permission

Alexander Oelzant aoe at
Sat Jul 29 22:05:58 GMT 2000

On Thu, Jul 27, 2000 at 02:26:24PM +0200, Urban Widmark wrote:
> On Thu, 27 Jul 2000 oliver.amft at wrote:
> Alexander Oelzant has a "symlink hack" here:
> This patch will probably not apply cleanly to a 2.2.13 kernel source tree
> (and if it did it may not work). If you want to test this it is safest to
> do so with the versions the patch was made for (2.3.99 -> 2.4.0-test4?).

This is correct, however, I've meanwhile "combined" the 2.0 and the 2.3
versions to a 2.2.(15) version, which seems to work fine. I haven't
tested it extensively, but apart from some harmless invalidation problems 
applications don't even see (only the kernel logs them, and there might
be a small performance hit), the hack seems about as stable as the 
other 0.6 versions.

For additional protection against accidentally set SYSTEM bits, symlimks 
now have a fixed length and the content is padded with zeroes up
to this file size.

> One thing that might be neat to do is to make the symlink format
> compatible with the format that the cygwin tools on 95/NT uses (I don't
> remember how they did it but it is similar, special permission & magic
> value). That would allow you to make a symlink on a mount and by using the
> cygwin tools you would be able to see it on the server too.

They rely on a cookie in the file header (supported by a system bit, too), 
and I must confess I am slightly scared of having to parse every file
that has the system bit set (both of the programming and the readdir
overhead).  However, if demand exists, I might some day make it compile
time or load time option. 

Enrico Weigelt also suggested using win9x style LNK files; since
Cygnus didn't go that way either, I don't see much sense in doing
it either (lose95 compatibility is not my major concern), besides
even the name would have to be faked somewhere in dir.c. yuck. 

Send suggestions, cookies and pointers my way ;-)

Current Information on the Symlink Hack:
2.0.33, 2.2.15 and 2.4.0-pre4 versions are available from
(with some description)
(without description)


ps. The CIFS-Enhancements from
look pretty interesting, except nothing has been done about the kernel modules
since 2.0 days and I cannot use the stuff anyway since the Samba server
we're connecting to isn't ours and patches are unlikely to be accepted

Alexander Oelzant 		alexander at
Durchlaufstr. 7/4/5		aoe at
A-1200 Wien			aoe at

More information about the samba mailing list