|
On Thursday 28 August 2008, Matthew Woehlke wrote: > Michael Howell wrote: > > On Thu, Aug 28, 2008 at 4:24 PM, Matthew Woehlke < > > > > mw_triad@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > >> Right, and since the other link may be in a different mountpoint (think > >> bind mounts or NFS), trying to find the other link can be difficult or > >> even impossible. > > > > No. Hard links must be on the same file system, since inode references > > cannot go cross file system. > > I didn't say anything about cross-filesystem... but bind mounts can put > one hunk of an FS in one place, and another somewhere else (and I don't > know of any reason a hardlink couldn't span this since it's still the > same physical FS). Similar with NFS mounts, the 'same filesystem' would > apply, I assume, the the server, which need not export the entire FS. bind and NFS is more like a symlink than a hard link (think of the mount point as a symlink to the destination FS). Both bind and NFS do not actually create new file name entries pointing to an inode on disk, therefore they are not hard links. hard links to not *have to* be on the same filesystem but I don't know of any POSIX implementations that allow cross-filesystem hard links, and POSIX allows implementations to forbid this. The reason you can't find the other links is that an inode is pointed to by file name entries but does not point back. i.e.: /a/file -> 654621 /b/file -> 654621 but knowing just inode 654621 is not enough to get back to /a/file and /b/file. Regards, - Michael Pyne |
Attachment:
signature.asc
Description: This is a digitally signed message part.
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<