Re: Unable to fsck 4.5TB FFSv2 fs

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

On Thu, 25 Sep 2008, Juergen Hannken-Illjes wrote:
> On Thu, Sep 25, 2008 at 04:58:46PM +0100, Stephen Borrill wrote:
> > On Thu, 25 Sep 2008, Christos Zoulas wrote:
> >
> > > On Sep 25, 12:15pm, netbsd%precedence.co.uk@localhost (Stephen Borrill) wrote:
> > > -- Subject: Re: Unable to fsck 4.5TB FFSv2 fs
> > >
> > > | ** /dev/rld1d
> > > | ** Last Mounted on /usr/backup
> > > | ** Phase 1 - Check Blocks and Sizes
> > > | cannot alloc 4294967292 bytes for inoinfo
>
> Just tried it at home on a -current i386 system as of early september.  As I
> don't have a 4.5TB disk here used a plain file (-F) on an UFS2 file system (*1):
>
> $ newfs -O 2 -f 4096 -b 32768 -I -F -s 9765867520 x3x
> x3x: 4768490.0MB (9765867520 sectors) block size 32768, fragment size 4096
>         using 6424 cylinder groups of 742.41MB, 23757 blks, 46848 inodes.
> super-block backups (for fsck_ffs -b #) at:
> 192, 1520640, 3041088, 4561536, 6081984, 7602432, 9122880, 10643328, 12163776,
> ...............................................................................
>
> $ fsck_ffs -F -f x3x
> ** x3x
> ** File system is already clean
> ** Last Mounted on
> ** Phase 1 - Check Blocks and Sizes
> ** Phase 2 - Check Pathnames
> ** Phase 3 - Check Connectivity
> ** Phase 4 - Check Reference Counts
> ** Phase 5 - Check Cyl groups
> 1 files, 1 used, 1201821133 free (13 frags, 150227640 blocks, 0.0% 
> fragmentation)
>
> Same result with newfs/fsck_ffs binaries from 4.0 running on -current.

Snap, I get the same on a fresh filesystem:

backup 14# fsck_ffs -F -f /dev/rld1d
** /dev/rld1d
** File system is already clean
** Last Mounted on
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
1 files, 1 used, 1201821133 free (13 frags, 150227640 blocks, 0.0% 
fragmentation)

> Are you sure your device is >4TB clean?

That's a good question. I added support for >2TB arrays as detailed here:
http://mail-index.netbsd.org/current-users/2008/09/16/msg004652.html

The changes were inspired by FreeBSD and I couldn't see anything obviously 
missing:
http://docs.freebsd.org/cgi/getmsg.cgi?fetch=351693+0+archive/2008/cvs-src/20080330.cvs-src

Any ideas of a good test case?

> Is there some magic using partition d (whole disk)?

I wouldn't have thought so, that's mainly to bypass any disklabel oddities 
without resorting to wedges.

--
Stephen