[PATCH 6/6] net: mvpp2: convert FDT access to ofnode API

[Simon Glass]

Hi Peng,

On 2026-05-26T07:25:09, Peng Fan (OSS) <peng.fan at oss.nxp.com> wrote:
> net: mvpp2: convert FDT access to ofnode API
>
> Convert mvpp2 driver from legacy fdtdec/fdt_* APIs to the ofnode-based
> interfaces.
>
> Replace usage of dev_of_offset(), fdtdec_lookup_phandle(),
> fdtdec_get_int(), fdt_parent_offset(), and related helpers with their
> ofnode equivalents, including dev_ofnode(), ofnode_parse_phandle(),
> ofnode_read_s32_default(), ofnode_get_parent(), and
> ofnode_for_each_subnode().
>
> Remove direct dependencies on gd->fdt_blob.
>
> Main changes:
> - Use ofnode_valid() instead of integer checks for node presence
> - Switch fixed-link detection to ofnode_find_subnode()
> - Replace uclass_get_device_by_of_offset() with
>   uclass_get_device_by_ofnode()
> - Update subnode iteration and device binding to use ofnode
>
> [...]
>
> drivers/net/mvpp2.c | 54 ++++++++++++++++++++++++++---------------------------
>  1 file changed, 27 insertions(+), 27 deletions(-)

Reviewed-by: Simon Glass <sjg at chromium.org>

> diff --git a/drivers/net/mvpp2.c b/drivers/net/mvpp2.c
> @@ -5436,15 +5436,14 @@ static struct driver mvpp2_driver = {
>   */
>  static int mvpp2_base_bind(struct udevice *parent)
>  {
> -     const void *blob = gd->fdt_blob;
> -     int node = dev_of_offset(parent);
> +     ofnode node = dev_ofnode(parent);

'node' is now no longer used by the look of it. Do you see a warning?

> diff --git a/drivers/net/mvpp2.c b/drivers/net/mvpp2.c
> @@ -5478,8 +5477,9 @@ static int mvpp2_base_bind(struct udevice *parent)
>               sprintf(name, 'mvpp2-%d', id);
>
>               /* Create child device UCLASS_ETH and bind it */
> -             device_bind(parent, &mvpp2_driver, name, plat,
> -                         offset_to_ofnode(subnode), &dev);
> +             ret = device_bind(parent, &mvpp2_driver, name, plat, subnode, &dev);
> +             if (ret)
> +                     return ret;

This is a behavioural change - the old code ignored bind failures and
continued, the new code aborts and leaks the already-allocated plat
and name. Probably the right thing, but please mention it in the
commit message (the "No functional changes" claim is not quite
accurate) and free plat/name on the error path.

> diff --git a/drivers/net/mvpp2.c b/drivers/net/mvpp2.c
> @@ -4731,33 +4731,34 @@ static int mvpp2_port_init(struct udevice *dev, struct mvpp2_port *port)
> -             parent = fdt_parent_offset(gd->fdt_blob, phy_node);
> -             ret = uclass_get_device_by_of_offset(UCLASS_MDIO, parent,
> -                                                  &port->mdio_dev);
> +             parent = ofnode_get_parent(phy_node);
> +             if (!ofnode_valid(parent))
> +                     return -ENODEV;

We should be careful returning -ENODEV as it has a special meaning
with DM. This is a small behavioural change versus the old code, which
passed a negative offset straight through. Worth noting in the commit
message, or drop the check since the call below will return an error
anyway.

Regards,
Simon