Re: [PATCH] Fix when installing on pationless but partionable medium

[Robert Millan]

On Fri, Jul 17, 2009 at 05:51:40PM +0100, Colin Watson wrote:
> On Fri, Jul 17, 2009 at 06:41:59PM +0200, Vladimir 'phcoder' Serbinenko wrote:
> > Sometimes a media that can be partioned isn't really partioned. E.g.
> > usb sticks. This is a patch to handle this situation. Unfortunately
> > such medium is often formated with a flavour of FAT which shares its
> > signature with MBR so it may be easily misidentified as
> > pc_partition_table. Furthermore the same signature is shared with
> > bootsectors including grub. One possibility is to try interpret disk
> > as known filesystems and see if we succeed. But the problem is that
> > the check for FAT are light and may result in false positives too. The
> > only more or less advanced check there is a check for FATXX string.
> > But I was about to propose to eliminate this check since I encountered
> > a FAT filesystem without this string on friend's SD card formatted
> > with symbian which he wanted to use as liveusb. Does anyone has a
> > better idea?
> 
> When checking for an MBR filesystem label, parted checks whether each of
> partitions 1-4 has a boot indicator that's either 0 or 0x80, since as
> you point out checking for the FAT signature suffers false positives; I
> believe this algorithm matches that in the Linux kernel. Look at
> libparted/labels/dos.c:msdos_probe(), which is already FSF-copyrighted
> and GPLv3+. GRUB should use the same algorithm, and then the worst case
> is that things will fail consistently.

I might be missing something about this check, but GRUB doesn't require that
the boot flag is present.  Therefore, its non presence doesn't imply this is
not a real msdos label.

-- 
Robert Millan

  The DRM opt-in fallacy: "Your data belongs to us. We will decide when (and
  how) you may access your data; but nobody's threatening your freedom: we
  still allow you to remove your data and not access it at all."