On Mon, 2001-12-03 at 13:59, D. Stimits wrote:
> Steve Lord wrote:
> >
> > It has always been possible for inode numbers in XFS to be more than 32
> > bits in size. This creates large problems for the rest of linux. This
> > change restricts the allocation of inodes in an xfs filesystem to that
> > part of the fs which will keep the inodes within 32 bits (basically
> > the bottom 1 Tbyte with default mkfs parameters). For these large
> > filesystems it also makes xfs file data prefer being higher up the
> > filesystem to retain space for inodes and metadata in the lower
> > portion.
> >
> > There is no on disk format change for this, but you will need to rebuild
> > mkfs to build a filesystem which can use it - there is some automatic
> > inode sizing code in mkfs which worked around the problem.
>
> Is there any chance a current filesystem on x86 2.4.x linux, disk sizes
> under 100 GB, would already have nodes greater than 32 bit? I'm assuming
> that upgrading won't break existing filesystems, i.e., do I need to fear
> using the 32 bit restrictions by mixing it with my current filesystem?
> I'm assuming that any ordinary size system (less than 1TB) will never
> have an inode above the 32 bit boundary. If not, I'd love a tool to
> sample all inodes and report the highest inode number, along with a
> simple "Exceeds 32 bits" or not type statement.
>
> D. Stimits, stimits@xxxxxxxxxx
>
No, there is no chance - it takes a 1Tbyte filesystem for the overflow
to occur.
Steve
--
Steve Lord voice: +1-651-683-3511
Principal Engineer, Filesystem Software email: lord@xxxxxxx
|