Re: [PATCH v3 1/4] rust: debugfs: Bind DebugFS directory creation

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Fri, May 02, 2025 at 09:05:25AM +0200, Danilo Krummrich wrote:
> On Fri, May 02, 2025 at 09:00:07AM +0200, Greg Kroah-Hartman wrote:
> > On Fri, May 02, 2025 at 08:37:40AM +0200, Danilo Krummrich wrote:
> > > On Thu, May 01, 2025 at 10:47:41PM +0000, Matthew Maurer wrote:
> > > > +/// Handle to a DebugFS directory that will stay alive after leaving scope.
> > > > +#[repr(transparent)]
> > > > +pub struct SubDir(ManuallyDrop<Dir>);
> > > 
> > > I think it's not very intuitive if the default is that a SubDir still exists
> > > after it has been dropped. I think your first approach being explicit about this
> > > with keep() consuming the SubDir was much better; please keep this approach.
> > 
> > Wait, let's step back.  Why do we care about the difference between a
> > "subdir" and a "dir"?  They both are the same thing, and how do you
> > describe a subdir of a subdir?  :)
> 
> We care about the difference, because Dir originally had keep() which drops the
> Dir instance without actually removing it. For subdirs this is fine, since
> they'll be cleaned up when the parent is removed.

But does that mean a subdir can not be cleaned up without dropping the
parent first?  For many subsystems, they make a "root" debugfs
directory, and then add/remove subdirs all the time within that.

> However, we don't want users to be able to call keep() on the directory that has
> been created first, since if that's done we loose our root anchor to ever free
> the tree, which almost always would be a bug.

Then do a call to debugfs_lookup_and_remove() which is what I really
recommend doing for any C user anyway.  That way no dentry is ever
"stored" anywhere.

Anyway, if Dir always has an implicit keep() call in it, then I guess
this is ok.  Let's see how this shakes out with some real-world users.
We can always change it over time if it gets unwieldy.

thanks,

greg k-h