[linux-cifs-client] Re: [PATCH 08/10] cifs: remove cifs_readlink and CIFSSMBQueryReparseLinkInfo

[Steve French]

I am just trying to figure out what happens when we hit this - if we
have a bug, lets fix it.  I haven't tried this in a while.

On Wed, Apr 29, 2009 at 9:35 AM, Jeff Layton <jlayton at redhat.com> wrote:
> On Wed, 29 Apr 2009 09:27:55 -0500
> Steve French <smfrench at gmail.com> wrote:
>
>> On Wed, Apr 29, 2009 at 8:29 AM, Jeff Layton <jlayton at redhat.com> wrote:
>> > CIFSSMBQueryReparseLinkInfo is only called by cifs_readlink, and
>> > cifs_readlink is never called by anything. A quick look at
>> > CIFSSMBQueryReparseLinkInfo shows that it has several problems
>> > with handling of endianness as well.
>>
>> What happens when we read a Junction with current code then?
>>
>> Windows allows users to create a "junction" (see the Windows utilities
>> "linkd.exe",  "ln.exe" and "fsutil").  Junctions look a little like a
>> symlink.  Although usually they have absolute paths, I think that they
>> can be created with relative paths like a symlink.    We probably
>> ought to make sure we can handle this before we remove this.
>>
>>
>
> Why? This code is officially unused since there's no way to call into
> it. I need to change the calling conventions for the unix symlink
> function, and this code also calls that. I don't really think my time
> is well spent fixing a dead code path.
>
> If/when you decide to fix the junction parsing code, this patch could
> be reverted. Until then, keeping this code around bloats out cifs.ko
> and adds to the maintenance burden with no benefit.
>
> --
> Jeff Layton <jlayton at redhat.com>
>



-- 
Thanks,

Steve