[conspire] question regarding mounting the Seagate FreeAgent Go 1T drive

Rick Moen rick at linuxmafia.com
Sun Mar 28 16:18:14 PDT 2010

Quoting Darlene Wallach (freepalestin at dslextreme.com):

> After I partition, format and label the Seagate FreeAgent Go 1T drive
> and I'm ready to mount it is this what I should use:
> mount -t fuseblk -o rw,nosuid, nodev,allow_other,blocksize=4096
> /dev/sdb1 /media/<myLabel>

Er, no.  You seem to have been somewhat lead astray by Ubuntu's
automouter into thinking that partition type "fuseblk" is somehow
necessary and desirable for this drive.

The latest and nicest NTFS driver for Linux, and also the initial
implementation of the Solaris ZFS filesystem on Linux, were developed
within the FUSE framework (a framework for mounting/umounting
filesystems strictly in userspace, which is not the standard Unix
method, which by contrast does the job in kernelspace).  Thus, the
Ubuntu automounter sniffed your terabyte USB drive, noticed a (default)
NTFS partition on it, and invoked the FUSE software to mount it.

That doesn't mean it's in your interest to continue to use the FUSE
layer after blowing away the NTFS and putting ext3 in its place.

FUSE is a nice idea in many ways:  For experimental code, you can play
with such things without putting something potentially dangerously buggy
into kernel space, where it would be much more likely to freeze your
system in a way that would necessitate hitting the power switch to
recover.  Also, in the case of ZFS, implementing the Linux filesystem
driver in userspace was a way of getting around the licensing conflict
that would apply for standard kernelspace code.  However, FUSE
filesystems are inherently a bit slower than kernelspace ones, and 
(for highly reliable filesystem code like ext3) it's highly in your
interest to avoid the overhead and additional complexity of relying on
the FUSE layer.

By the way, after you make the ext3 filesystem on your USB drive, I have
no idea if the Ubuntu automounter might decide to mount it without
consulting you.  (I tend to kill automounters with extreme prejudice
before I could learn what holes they're likely to shoot in my feet.)

I've never seen the allow_other mount option, before.  Looks handy.  My
Debian system's version of mount doesn't yet support it, so it must be
really new.  It _might_ be FUSE-only.

The blocksize=4096 option might also be FUSE-only.  In any event, with
regular kernelspace-based filesystems, you shouldn't need that.  You can
specify a block size when you run mkfs.ext3, but I would usually leave
that up to the utility's heuristics to pick on my behalf.  And, having
done that, I'm pretty sure you don't need to specify blocksize in the
mount command:  It figures it out.

Listen, if you dispose of the FUSE stuff, you get something like this:

  mount  -t ext3  -o nosuid,nodev   /dev/sdb1  /media/<myLabel>

Once you've tested that, _if_ you're planning on leving this drive
connected to the machine most/all of the time, then add a line to
/etc/fstab and let the startup routines take care of it for you.

If the drive will be sometimes connected, sometimes not, you can still
add a line to /etc/fstab but include the noauto option.  This results in
the drive _not_ being mounted at startup time, but makes it a lot less
tiresome to manually mount it from the command line:

  sudo mount /dev/sdb1

Because of the line in /etc/fstab, the "mount" command looks up your
preferred options, mountpoint, etc., from there.

> Should I add - uhelper=hal to the mount?

{shrug}  I tend to stay as far away from HAL stuff as I can, but then
I'm a certifiable fossil who simply hasn't warmed to a lot of the stuff
comming out of the freedesktop.org / GNOME people.  So, again, suit

I'd never even heard of that mount option until you mentioned it.
Searching turns up:

> mount -t fuseblk -o rw,nosuid,
> nodev,uhelper=hal,allow_other,blocksize=4096 /dev/sdb1 /media/<myLabel>
> what about adding shortname=lower to the mount?

That's a new one to me, too -- but I'm guessing it's only for FAT
filesystems.  ext3 is a real filesystem, and case-sensitive by design.

More information about the conspire mailing list