Mount and tune2fs disagree

[Luciano Miguel Ferreira Rocha]

On Sun, Apr 04, 2004 at 03:12:59PM +0200, shmuel siegel wrote:
> On Sun, 2004-04-04 at 14:47, Luciano Miguel Ferreira Rocha wrote:
> > On Sun, Apr 04, 2004 at 02:38:39PM +0200, shmuel siegel wrote:
> > > Thank you. Mounting as ext3 was sufficient. I would have done this had
> > > the default mount failed. It didn't dawn on me that mount could
> > > legitimately mount the file system "wrong". Does some advantage accrue
> > > to a fs layout that permits this "error"?
> > 
> > Advantage? No, I don't think so. Had you created any files on the FAT
> > partition, you'd probably corrupt the ext3 one.
> > 
> > The problem is that the mount command had no way to know that, although
> > the partition was successfully mounted as vfat, it shouldn't be.
> > 
> This last statement, I disagree with. FDISK recognizes the partition as
> a linux partition, parted says that it is an ext3 partition, and, IIRC,
> partition magic also recognizes it as an ext3 partition. So there seems
> to be enough information running around to properly identify the
> partition. But at least now I understand why Windows thought that it
> could mount the partition but always got the capacity wrong.
> The next time I get around to booting Windows, I will check out what
> partition magic thinks. If it does recognize the partition as ext3, I
> will enter this information into bugzilla.

But the partition type does not enforce a partition filesystem. You can
put any filesystem in any partition, regardless of its type.

The rationale for partition types is the same for file extensions. It's
easier to consult a table with (id, type) relations than to parse the
filesystem superblock.

Then we end with 6 different types for FAT partitions, and several
different interpretations for .doc files...

But back on topic, the reasons for not using the partition type by mount
to find the filesystem type, is that the partition type isn't available
for userland programs (at least, I don't know any besides reading the
MBR), and some times it isn't a partition (floppies, cdroms, loop
devices).

And some times, the partition type isn't the correct one. After all,
Windows did mount the partition, didn't it?

So, personally, I don't think it's a bug. The irrelevance of filesystem
types in /etc/filesystems is, if it's still in the Fedora Core test
releases.

Regards,
Luciano Rocha