Re: Symlinks on Windows-based shares

On Fri, Jan 08, 2010 at 11:23:35AM -0500, Jeff Layton wrote:
> The tricky part here is that that would require some overhead to
> determine whether an inode is one of these symlinks. Today, if I stat a
> file, I do a QPathInfo call (similar to a GETATTR on NFS) and maybe a
> second QPathInfo call with a different infolevel to get the unique id.
> 
> With that scheme, you'll be turning each stat() call of a file into a
> read operation. That requires opening and closing the file too. Each
> stat() operation would go from 1 or 2 calls on the wire to 4 or 5.
> 
> I'm not opposed to a patchset that implements those, but would probably
> suggest that it be configurable via a mount option or something.

Good point, and a mount option seems like a perfect solution in this
case.  I'll see what I can come up with and get back to you guys, then.

--Doug