Re: [Avf-fuse-dev] Hi, fusermount problems and fuse - kioslave bridge :-)

[<Mik...@et...>]

> > It can work two ways:
> >
> >   1) It opens the fuse communication device, mounts a fuse filesystem
> >      and starts a filesystem daemon given on the command line passing
> >      it the fuse communication file descriptor as stdin (0)
> 
> does this mean I have to remove the fuse_mount() call and use 0 instead of 
> fuse_fd in the code you can see below ?

Yes, if you want to use this method, but as I said this is deprecated,
and I use it just for testing purposes.

> >   2) The filesystem daemon invokes fusermount which opens the fuse
> >      communication device, mounts the filesystem and passes the
> >      communication file descriptor to the invoking program through a
> >      UNIX domain socket.
> 
> Currently our "filesystem daemon" does the following:
> 
>    int fuse_fd = fuse_mount(mount_point.path().latin1(), 0);
>    fuse_handle = fuse_new(fuse_fd, 0, &ops);
>    if (fuse_handle==0)
> 	kdDebug()<<"fuse_new() failed"<<endl;
>    fuseApp = new FuseApp(base_url);
>    set_signal_handlers();
>    fuse_loop(fuse_handle);
>    close(fuse_fd);
>    fuse_unmount(mount_point.path().latin1());
> 
> Is this what you describe here ?

Exactly.

> I saw in an email from the mailing list that you just added an option
> >>>OK, so a new option for fusermount is needed, e.g.:
> >>>   fusermount -d xyz /mnt
> >>>will insert the following line into the mtab:
> >>>   xyz   /mnt  ...
> >>>Does this sound OK?
> 
> to fusermount. If our daemon works the way as it does now, is there a way to 
> use this option ?

You can use the 'args' parameter of fuse_mount():

  const char *args[] = { "-d", "xyz", NULL };
  int fuse_fd = fuse_mount(mount_point.path().latin1(), args);  

This should work (though not tested).

> When I run our daemon and mount an ioslave, the daemon exits with an error 
> message when I umount it with fusermount -u mountpoint. Is this ok ?

Yes.  I'll look at the code to see if this can be improved.

Miklos