New fusefs implementation not usable with multiple fusefs mounts

[Kevin Oberman]

On Tue, Jul 2, 2013 at 3:20 PM, Attilio Rao <attilio at freebsd.org> wrote:

> On Wed, Jul 3, 2013 at 12:07 AM, Kevin Oberman <rkoberman at gmail.com>
> wrote:
> > I have been using the new fusefs for a while and have had to back it out
> and
> > go back to the old kernel module. I keep getting corrupted file NTFS
> systems
> > and I think I understand why,
> >
> > I mount two NTFS systems:
> > /dev/fuse    184319948 110625056 73694892    60%    /media/Media
> > /dev/fuse    110636028 104943584  5692444    95%    /media/Windows7_OS
> >
> > Note that both systems are mounted on /dev/fuse and I am assured that
> this
> > is by design. Both work fine for reads and seem to work for writes. Then
> I
> > unmount either of them. Both are unmounted, at least as far as the OS is
> > concerned. There is no way to unmount one and leave the other mounted. It
> > appears that any attempt to unmount either system does a proper unmount
> of
> > /media/Media, but, while marking /media/Windows7_OS as unmounted,
> actually
> > does not do so. The device ends up corrupt and the only way I have been
> able
> > to clean it is to boot Windows and have a disk check run. Media never
> seems
> > to get corrupted.
> >
> > Any further information I might gather before filing a PR? I am running
> on
> > 9.1 stable, but havehad the problem since the patch set first became
> > available on 9.0-stable.
>
> I do not understand, new fusefs implementation was never committed to
> stable branch to my knowledge.
> Did you backport manually?
>
> BTW I cc'ed George which should maintain the module.
>
> Attilio
>

Attilio,

Actually, you provided the patches for 9-Stable way back when you first did
them and we had an exchange on current@ about their use on 9-stable and
their operation including the mounts all being on /dev/fuse. I also edited
the mount_fuse man pages to clarify the awkward wording of the original
(which you didn't write).

They still apply pretty cleanly and I continued using them until about 3
weeks ago when I removed them to test whether they were responsible for the
issues I was seeing. Since I got corruption most every time I unmounted the
file systems after having written to the Windows one, I am now pretty sure
that it does not happen when I use the old kernel module.

The analysis of the problem is purely speculation, but fits the behavior.
If it is correct, I would expect the same issues to occur with head.

Thanks for copying George. I didn't realize that he had taken over the
code. I won't bu you about it again.
-- 
R. Kevin Oberman, Network Engineer
E-mail: rkoberman at gmail.com