Re: Fatal trap 12: page fault while in kernel mode (FUSE related?)

On Fri, Nov 1, 2013 at 6:03 AM, Alexey Dokuchaev <danfe_at_nsu.ru> wrote:

> On Thu, Oct 31, 2013 at 09:59:42AM -0700, Kevin Oberman wrote:
> > On Tue, Oct 29, 2013 at 3:24 AM, Alexey Dokuchaev <danfe_at_nsu.ru> wrote:
> > > I was running out of space on my UFS partition and decided to use big
> NTFS
> > > one I also have on the drive.  I've mounted it with ntfs-3g and our
> native
> > > fuse.ko.  I needed the scratch space to built Open/LibreOffice on it
> *LOL*.
> > > Well, it failed with a panic (see the excerpt from text core at the
> end of
> > > this email; full debug info is available upon request).
> >
> > I get a very similar panic when I attempt an rsync from a remote system
> > to my NTFS drive. Very easy to reproduce. Something in fuse goes off the
> > rails under active R/W activity, it seems.
>
> Hmm, given more people are seeing it, and it's not too hard to reproduce,
> I hope it can be tracked down and nailed.  I will enable debugging features
> in my kernel so I can gather some data when this shit happens again to me.
>
> ./danfe
>

Just to be clear, my software was built from source, so the package repo is
irrelevant. Also, I am running 9.2-Stable with fuse in the kernel (not a
module) backported from the 10.0 code and the mount_fuse from 10.0, as
well, so mine is a rather odd system. I should be able to capture  dump,
but I don't have one now.

I should also mention that the FreeBSD kernel code was updated in July to
7.10, but was reverted five days later to 7.8 due to issues with
fusefs_libs not getting along with it. At the time it was suggested that
fusefs_libs would need to be updated to match. (I have no idea how tricky
this might be, but it is way beyond my capability.

Guess it's almost time to update to 10.0. I really should have moved to
HEAD before 10-STABLE was branched. Then I'll be running fully supported
code.
-- 
R. Kevin Oberman, Network Engineer
E-mail: rkoberman_at_gmail.com