Re: [PATCH 10/11] platform/x86: intel_cht_int33fe: Link with external dependencies using fwnodes

[Heikki Krogerus <heikki.krogerus@linux.intel.com>]

On Mon, Apr 08, 2019 at 07:42:52PM +0300, Andy Shevchenko wrote:
> On Wed, Apr 3, 2019 at 7:46 PM Heikki Krogerus
> <heikki.krogerus@linux.intel.com> wrote:
> >
> > Supplying also external devices - the DisplayPort connector
> > and the USB role switch - software fwnodes. After this the
> > driver has access to all the components tied to the USB
> > Type-C connector and can start creating software node
> > references to actually associate them with the USB Type-C
> > connector device.
> 
> > +static int role_switch_match(struct device *dev, void *name)
> > +{
> > +       return !strcmp(dev_name(dev), name);
> > +}
> > +
> > +static int cht_int33fe_setup_mux(struct cht_int33fe_data *data)
> > +{
> 
> > +       /* First let's find xHCI PCI device */
> > +       pdev = pci_get_class(PCI_CLASS_SERIAL_USB_XHCI, NULL);
> > +       if (!pdev || (pdev->vendor != PCI_VENDOR_ID_INTEL))
> > +               return -ENODEV;
> > +
> 
> > +       /* Then the child platform device */
> > +       p = device_find_child(&pdev->dev, "intel_xhci_usb_sw",
> > +                             role_switch_match);
> 
> We have already similar code in kernel
> (drivers/nvdimm/claim.c::namespace_match), perhaps
> you at least can (locally) introduce
> 
> static int match_name(...) // or re-use from drivers/base/bus.c
> {
> ...
> }

OK.

> static device_find_child_by_name(...)
> {
> ...
> }
> 
> which we may move to the driver core now or later on.
> 
> > +       pci_dev_put(pdev);
> > +       if (!p)
> > +               return -EPROBE_DEFER;
> > +
> 
> > +       /* Finally the mux device */
> > +       dev = device_find_child(p, "intel_xhci_usb_sw-role-switch",
> > +                               role_switch_match);
> 
> Ditto.
> 
> > +       put_device(p);
> > +       if (!dev)
> > +               return -EPROBE_DEFER;
> 
> > +       return 0;
> 
> > +}
> 
> > +static int cht_int33fe_setup_dp(struct cht_int33fe_data *data)
> > +{
> > +       struct fwnode_handle *fwnode = data->node[INT33FE_NODE_DISPLAYPORT];
> > +       struct pci_dev *pdev;
> > +
> > +       /* First let's find the GPU PCI device */
> > +       pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, NULL);
> > +       if (!pdev || (pdev->vendor != PCI_VENDOR_ID_INTEL))
> > +               return -ENODEV;
> > +
> > +       /* Then the DP child device node */
> > +       data->dp = device_get_named_child_node(&pdev->dev, "DD02");
> > +       pci_dev_put(pdev);
> > +       if (!data->dp)
> > +               return -ENODEV;
> > +
> 
> > +       fwnode->secondary = ERR_PTR(-ENODEV);
> > +       data->dp->secondary = fwnode;
> 
> Can't generic helper be in charge of doing this, i.e. chaining fwnodes?
> 
> ///
> void fwnode_chain(parent, new)
> {
> ...
> }
> ///

We probable should just export set_secondary_fwnode(). I'll add a
patch for that.


thanks Andy,

-- 
heikki