On Tue, 2002-04-30 at 18:49, Andi Kleen wrote:
> On Tue, Apr 30, 2002 at 02:31:15PM -0800, Ethan Benson wrote:
> > On Tue, Apr 30, 2002 at 11:27:06PM +0200, Andi Kleen wrote:
> > > On Tue, Apr 30, 2002 at 09:53:02AM -0500, Steve Lord wrote:
> > > > I said you cannot do it, I was referring to the existing code, yes it
> > > > is possible, but currently only ext2 and ext3 support this, and chattr
> > > > is an ext2 utility, not a generic linux filesystem utility.
> > > >
> > > > Adding it also introduces an on disk incompatibility between Irix and
> > > > Linux, and probably we could not take the code back to Irix, given its
> > > > origin. So we would need to version the superblock - we would need that
> > > > anyway for backward compatibility. All in all it gets to be a larger
> > > > project than you might think. My problem now is definitely not lack of
> > > > work ;-)
> > >
> > > If one would put it into a new extended attribute then I guess it could
> > > be compatible to irix.
> > >
> > > Disadvantage: checking it is more costly than just a bit in the main
> > > inode,
> > > but I guess when ACLs is enabled it has to search the EAs for each inode
> > > open anyways, so it probably won't make too much difference.
> >
> > i believe that would be considered too ugly of a kludge,
> > immutable/append-only are bits, not attributes. it also gets messy
>
> It's just an implementation detail how the bits are stored, users do not
> notice as long as the ext2 attr ioctl is supported.
>
> -Andi
I dug some more and there does not appear to be checking on unused bits
in the di_flags field of the on disk inode, although that does not
include xfs_check which is a rather byzantine chunk of code. So it
might be possible to use a bit in here. Like I said though, right
now I am not going to get near something like this for quite a while.
Andi, is immutable checking all done above the vfs or do filesystems
have to enforce it as well?
Steve
|