On Thu, Mar 23, 2023 at 02:29:24PM +0000, Russell King (Oracle) wrote:
> On Thu, Mar 23, 2023 at 03:59:07PM +0200, Andy Shevchenko wrote:
> > On Wed, Mar 22, 2023 at 11:59:55AM +0000, Russell King wrote:
> > > From: Vladimir Oltean <vladimir.oltean@xxxxxxx>
> > >
> > > Allow a named software node to be created, which is needed for software
> > > nodes for a fixed-link specification for DSA.
...
> > > +fwnode_create_named_software_node(const struct property_entry *properties,
> > > + const struct fwnode_handle *parent,
> > > + const char *name)
> > > {
> > > struct fwnode_handle *fwnode;
> > > struct software_node *node;
> > > @@ -930,6 +931,7 @@ fwnode_create_software_node(const struct property_entry *properties,
> > > return ERR_CAST(node);
> > >
> > > node->parent = p ? p->node : NULL;
> > > + node->name = name;
> >
> > The same question stays as before: how can we be sure that the name is unique
> > and we won't have a collision?
>
> This got discussed at length last time around, starting here:
>
> https://lore.kernel.org/all/YtHGwz4v7VWKhIXG@xxxxxxxxxxxxxxxxxx/
>
> My conclusion is that your concern is invalid, because we're creating
> this tree:
>
> node%d
> +- phy-mode property
> `- fixed-link node
> +- speed property
> `- full-duplex (optional) property
>
> Given that node%d will be allocated against the swnode_root_ids IDA,
> then how can there possibly be a naming collision.
>
> You would be correct if the "fixed-link" node were to be created at
> root level, or if we were intentionally creating two swnodes under
> the same parent with the same name, but we aren't.
>
> Plus, the code _already_ allows for e.g. multiple "node1" names - for
> example, one in root and one as a child node, since the code uses
> separate IDAs to allocate those.
>
> Hence, I do not recognise the conern you are raising, and I believe
> your concern is not valid.
>
> Your concern would be valid if it was a general concern about
> fwnode_create_named_software_node() being used to create the same
> named node under the same parent, but that IMHO is a programming
> bug, no different from trying to create two devices under the same
> parent with the same name.
>
> So, unless you can be more expansive about _precisely_ what your
> concern is, then I don't think there exists any problem with this.
OK.
I leave it to others to review. I have nothing to add.
--
With Best Regards,
Andy Shevchenko
