On Thu, Mar 02, 2000 at 10:19:40AM -0500, Allan M. Wind wrote:
> On 2000-03-01 23:42:37, kmself@ix.netcom.com wrote:
> > On Wed, Mar 01, 2000 at 07:05:04PM -0500, Allan M. Wind wrote:
> > > On 2000-03-01 23:42:21, Mary Honeycutt wrote:
>
> > tar cf - source | ( cd /target; tar xpf - )
>
> Yes, that would be more like it.
Thought so. I'd hate for someone to find out the hard way....
> > There's a utility to recreate the lost+found directory if you do manage
> > to overwrite it -- it needs to sit on a specific inode for the filesystem
> > to be able to recover lost clusters properly. RTFM, it's there somewhere.
>
> Isn't lost+found created by mkfs?
Yes, it is (or mke2fs, or whatever). My understanding of this is
somewhat limited, but here goes.
Files and directories are identified under most Linux-like fileystems
(e2fs, minix fs, UFS, etc., but *not* msdos, vfat), by inodes. An inode
is essentially a database entry in a table giving storage location,
name, and values of several attributes (read/write/execute/suid), etc.
When your disk gets f*cked, you fsck it. This does magic to inodes. (I
*said* my understanding was limited). When it can't figure out what
magic to do, like the name of the file was lost, it dumps the "lost"
inodes to a special directory. Because the *name* of that directory
might itself be lost, it *always* sits on the same inode of the
filesystem -- though this location varies with the specific filesystem
in question IIRC.
So you can't just "mkdir lost+found" and hope everything works.
man (8) mklost+found for more info.
BTW: Allen, you might want to fix your "Reply-to:" header.
--
Karsten M. Self (kmself@ix.netcom.com)
What part of "Gestalt" don't you understand?
Scope out Scoop: http://scoop.kuro5hin.org/
Nothin' rusty about Kuro5hin: http://www.kuro5hin.org/
