Hello,
On Tue, Jul 07, 2009 at 09:54:51PM +0200, Carl Fredrik Hammar wrote:
> On Tue, Jul 07, 2009 at 09:24:27PM +0300, Sergiu Ivanov wrote:
> > In the latest working design unionmount creates a proxy node (by
> > cloning the netfs_root_node of unionfs translator) and sets the
> > mountee on this proxy. I'm currently trying to implement cfhammar's
> > idea about having the mountee run in orphan mode. To achieve this I
> > call only fshelp_start_translator, with no file_set_translator
> > following. When calling fshelp_start_translator, I have to give a
> > pointer to a function, open_port, which in my case looks like this:
> >
> > /* Opens the port on which to set the mountee. */
> > error_t
> > open_port (int flags, mach_port_t * underlying,
> > mach_msg_type_name_t * underlying_type, task_t task,
> > void *cookie)
> > {
> > err = 0;
> >
> > /* Create a port to `np`. */
> > newpi = netfs_make_protid
> > (netfs_make_peropen (np, flags, NULL), user);
> > if (!newpi)
> > {
> > iohelp_free_iouser (user);
> > return errno;
> > }
> >
> > *underlying = underlying_port = ports_get_send_right (newpi);
> > *underlying_type = MACH_MSG_TYPE_COPY_SEND;
> >
> > ports_port_deref (newpi);
> >
> > return err;
> > } /*open_port */
> >
> > np is the pointer to the proxy node. If I want to get rid of the
> > proxy node I must somehow avoid keeping references to it. However, in
> > the above code I clearly add a reference to the proxy node by creating
> > a port which goes to the mountee and thus, does not get destroyed
> > immediately.
>
> Is there any reason to have a proxy at all now that there's no
> file_set_translator? Why not simply pass unionfs's underlying node?
My previous mail says that when I simply pass the unionfs's underlying
node I get an error; the error is given in that mail :-) That's why
the question in the subject of the post ;-)
Regards,
scolobb
