Re: [fuse-devel] Auto unmount on segfault

[Kevin F. <Kev...@pn...>]

I think that is somewhat file system specific...

In some cases, you may want an unmount.

In others, for example, if you are rsyncing with --delete, you really
really want a file system failure to give you errors, not a blank
directory.

Kevin

On Thu, 2010-09-09 at 09:47 -0700, Max Krasnyansky wrote:
> On 09/09/2010 06:37 AM, Stef Bon wrote:
> > Sorry for my last mail. It's not answering your question at all.
> It's ok :).
> 
> > I do not know how to deal with segfaulting. Just take care it does not
> > segfault, but you'll probably know that.
> > Isn't there a "on segfault do this and that" call?
> I guess segfault is just an example. "kill -9" would be another one.
> In other words any unclean shutdown of the process running FUSE.
> And no, catching a segfault is not an option (I mean you can of course 
> but I'd rather not, and you can't catch SIGKILL).
> What I want is the same behavior as the file descriptors - process dies 
> FDs get closed and things get cleaned up. One way to implement that is 
> to have a helper process that waits for parent to die and cleans up, and 
> I'm thinking that fusermount could be that process.
> 
> Max
> 
> 
> >
> > Stef
> >
> > 2010/9/9 Stef Bon<st...@gm...>:
> >> 2010/9/9 Max Krasnyansky<ma...@qu...>:
> >>> On 08/27/2010 01:14 PM, Max Krasnyansky wrote:
> >>>> Hi Folks,
> >>>>
> >>>> I did some searching and reading of the FAQ&    list archive and I cannot
> >>>> seem to find any prev discussions on this. Let me know if I missed some.
> >>>>
> >>>> My use case is pretty simple. An app exposes some internal stuff via
> >>>> FUSE (ala /proc and /sys fs). The fs is lazy unmounted when the app
> >>>> exits cleanly. ie it shuts down FUSE and calls the unmount. However if
> >>>> the app dies (segfault for example) the mount stays and returns errors
> >>>> on access.
> >>>> So the question is what's the best way to do the auto lazy unmount in
> >>>> that case. Initially I was going to fork() a child and have the child
> >>>> wait for the parent to die and then do the unmount. But now I'm thinking
> >>>> that since we already calling 'fusermount' why not a
> >>>> --unmount-on-parent-death option to fusermount.
> >>
> >> Hmm unmount on parent death?? Sounds like multithreading. Is single
> >> mode no option?
> >>
> >> It's not a must to use multithreading.
> >>
> >> Stef
> >>
> 
> 
> ------------------------------------------------------------------------------
> This SF.net Dev2Dev email is sponsored by:
> 
> Show off your parallel programming skills.
> Enter the Intel(R) Threading Challenge 2010.
> http://p.sf.net/sfu/intel-thread-sfd
> _______________________________________________
> fuse-devel mailing list
> fus...@li...
> https://lists.sourceforge.net/lists/listinfo/fuse-devel