Re: Problem mounting new Sandisk 1TB USB drive

Late to the 'old and stupid' party ... out of the pile, I'm responding to this one since Gary gives it as reference in his PR revival at
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=244461

On 11 January 2024 11:42:15 am AEDT, Polytropon <freebsd@edvax.de> wrote:
 > On Wed, 10 Jan 2024 16:42:22 -0700, Gary Aitken wrote:
 > > On 1/10/24 15:33, Polytropon wrote:
 > > > On Wed, 10 Jan 2024 13:03:28 -0700, Gary Aitken wrote:

(cutting mercilessly)

 > > > The stick is actually pre-formatted with exFAT.
 > > > 
 > > > Problem here: exFAT disguises itself as an NTFS partition,
 > > > but it does not have NTFS metadata and content, so ntfs-3g
 > > > cannot mount it, even though tools like gpart or fdisk
 > > > show a NTFS partition.

It's worse than that, thanks again Microsoft!  Reference: https://en.m.wikipedia.org/wiki/Partition_type

MBR partition type 7 can be IFS, HPFS (both OS/2), NTFS or ExFAT (both MS).

It's kinda understandable between HPFS and NTFS; well before FUSE there was read safely / write maybe code for both that was largely identical. I used it to read old OS/2 disks up to about FreeBSD 9.

But overlaying NTFS with ExFAT was unforgivable really ... as this thread adequately demonstrates.

 > > > So here is what you should check:
 > > > 
 > > > 1. /boot/loader.conf contains:
 > > > 
 > > > 	fuse_load="YES"

 > > # Enable FUSE functionality for exfat filesystem mounting
 > > fuse_enable="YES"
 > > fuse_load="YES"
 > 
 > THe _enable settings belong to /etc/rc.conf; /boot/loader.conf
 > uses the _load settings.

 > > > 2. Mount device manually, read-only, perform checks, then
 > > >     unmount again:
 > > > 
 > > > 	# mount.exfat -o ro /dev/da0s1 /mnt/memstick
 > > > 	# df -f /mnt/memstick
 > > > 	# ls -R /mnt/memstick
 > > > 	# umount /mnt/memstick
 
 > Should have been "df -h". :-)
 > 
 > Intention: Compare "dmesg" entry (device reporting size)
 > with filesystem size, and then compare to what's written
 > on the device itself.
 
 > > A series of mount.exfat / umount / mount.exfat on fbsd seemed to
 > > clear it up.

 > There are also fsck tools for most filesystems, and there
 > probably is one for exFAT, just in case.

sysutils/exfat-utils : exfatfsck(8)
also mkexfatfs(8), dumpexfat(8), exfatlabel(8).

 > > > NB: mount.exfat does not have a manpage or help option.
 > > 
 > > $ apropos exfat
 > > mount.exfat-fuse(8) - mount an exFAT file system

 > Okay, thanks for that pointer. It's not obvious that you
 > need the manpage of "mount.exfat-fuse" when you want to
 > know the options of "mount.exfat" though...

sysutils/fusefs-exfat is installed as mount.exfat-fuse, with a symlink for /usr/local/sbin/mount.exfat

Unfortunately there's no such link for manpage mount.exfat-fuse(8)

 > > > Why don't you simply run newfs on the stick, with the
 > > > filesystem *you* _want_ to use? You can use UFS it you
 > > > want to, works perfectly well. ;-)
 > > 
 > > I want to be able to read it from an android phone, and I was/am
 > > concerned reformatting may make it unreadable by the phone.

I was pleasantly surprised to hear that Android reads/writes exfat. I installed it for Ventoy, but can now directly write from the phone, ta! I also have some 64G USB2 sticks that become useful with exfat.

 > That's a trap to easily fall into: exFAT does not identify
 > itself as exFAT or FAT, but as NTFS, but it is not NTFS.

Nor is it HPFS (in about 7 years of OS/2 in the nineties, I can't recall losing a single file to HPFS)

~~~~~~~

To address points Gary raises in the abovementioned PR:

-> I would consider it a bug. Documenting ambiguous behavior does not make it any less of a problem.

Agreed.

-> It affects anyone who attempts to mount a usb stick with >32GB of storage.

Well, and smaller exfat media too.

-> If gpart can tell that a partition is exFAT, it should report it as such to avoid confusion

It doesn't seem like a small 'if'.

It means adding code to detect the real type in use - and there may be none before a makefs.

And it's not binary; IFS and HPFS as well.  Maybe there's some simple test to delineate between those four, some clear signature?  I don't know; you'd need to examine fsck (or analogous) code, test and present it to the gpart maintainer.

Good luck!