Re: grub-probe detects ext4 wronly as ext2

[Javier Martín]

El mié, 16-07-2008 a las 19:44 +0200, Felix Zielcke escribió:
> > Oh well I should have used grep with -i
> > meta_bg and META_BG does make a difference
> >
> > Anyway in release-notes I now found this:
> >
> > Add support for the an alternative block group descriptor layout which
> > allows for on-line resizing without needing to prepare the filesystem
> > in advance.  (This is the incomat feature flag meta_bg.)
> >
> > So it seems you'll have sooner or later to worry about that.
> > But I still think it's currently not that important to implement it, 
> > because it's currently not compatible with resize_inode
> >
> 
> I should really take me more time for such things, but I think it's at least 
> better then sending the same mail twice :)
> (Robert Millan knows why I say this)
> 
> Anyway I thought that the release-notes are just for the current release not 
> complete list for every like NEWS or even the changelog
> This entry is from version 1.30 dated October 31, 2002
> So I think resize_inode has completly replaced meta_bg
> It's at least not mentioned anymore in the mke2fs(8) man page or even set by 
> mke2fs.conf
> But probably the question remains, if grub should ignore it or reject it if 
> it's set

OK, so this is what I get from your 3 posts, and my proposals for the
driver future:
        * meta_bg is a deprecated feature and is incompatible with
resize_inode, which is used by default in ext4 and can be used by all
other extN filesystems. We cannot safely ignore it since it signals an
incompatible change in one of the fs descriptors, so the reading code
might not be able to locate (meta)data on disk. I advocate _rejecting_
volumes with it (given that it's no longer being generated by any
"standard" tool) until someone writes a patch to support it.
        * flex_bg seemingly allows certain metadata structures to be located at
places other than their default positions. However, given that it is
only used on volumes quite filled with bad blocks, I think we can (at
least temporarily) _ignore_ it, but truly supporting it would be a Good
Thing (tm)
        * uninit_bg might already be supported (we should check, though, with
bigger partitions) and is not an incompat flag, so my patch does not
affect its handling.

signature.asc
Description: Esta parte del mensaje está firmada digitalmente