On Tue, Jul 22, 2003 at 01:55:27PM +1000, Ivan Rayner wrote:
> On Mon, 21 Jul 2003 16:34:43 -0800, Ethan Benson wrote:
>
> > However during my testing I did discover 3 issues:
> ...
> > 2: xfsrestore breaks when immutable or append-only files are included
> > in the dump. The problem is xfsrestore restores the file inode flags
> > too soon, and then loses the access it needs to restore extended
> > attributes, uid, gid, and permissions. I am not familiar enough with
> > xfsrestore to really propose the fix, however i believe just changing
> > it to restore at the very least immutable/append-only flags very last
> > after all other restoration is complete should suffice to solve this
> > problem.
>
> Your explanation makes sense. xfsrestore, at the moment, will restore
> extended attributes (EAs) after all of the data for the file has been
> restored. This is important for DMF, where it would be a bad idea for DMF
> to start operating on a file before it was totally restored because the
> relevant EA was restored too early. (Tim fixed a bug related to this some
> time ago.)
>
> It probably wouldn't be too difficult to fix xfsrestore to handle the new
> flags in the same way as extended attributes. ('Course that's easy to say
> when you haven't looked at any of the code in question ... :)
it didn't look too difficult to me to change, but I am hesitent to
really make the attempt since there are things about
xfsdump/xfsrestore for which I don't know (such as ordering issues you
describe above). I also don't really want to put much effort into it
until the kernel code is actually accepted.
--
Ethan Benson
http://www.alaska.net/~erbenson/
pgp8FDsmUNi2i.pgp
Description: PGP signature
|