2024-05-03 20:01:38

by Hiago De Franco

[permalink] [raw]
Subject: [PATCH] net: ethernet: ti: am65-cpsw-nuss: create platform device for port nodes

From: Hiago De Franco <[email protected]>

After this change, an 'of_node' link from '/sys/devices/platform' to
'/sys/firmware/devicetree' will be created. The 'ethernet-ports' device
allows multiple netdevs to have the exact same parent device, e.g. port@x
netdevs are child nodes of ethernet-ports.

When ethernet aliases are used (e.g. 'ethernet0 = &cpsw_port1' and
'ethernet1 = &cpsw_port2') in the device tree, without an of_node
device exposed to the userspace, it is not possible to determine where
exactly the alias is pointing to.

As an example, this is essential for applications like systemd, which rely
on the of_node information to identify and manage Ethernet devices
using device tree aliases introduced in the v251 naming scheme.

Signed-off-by: Hiago De Franco <[email protected]>
---
drivers/net/ethernet/ti/am65-cpsw-nuss.c | 7 +++++++
1 file changed, 7 insertions(+)

diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
index 1d00e21808c1..f74915f56fa2 100644
--- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
+++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
@@ -2091,6 +2091,13 @@ static int am65_cpsw_nuss_init_slave_ports(struct am65_cpsw_common *common)
if (strcmp(port_np->name, "port"))
continue;

+ if (!of_platform_device_create(port_np, NULL, NULL)) {
+ dev_err(dev, "%pOF error creating port platform device\n",
+ port_np);
+ ret = -ENODEV;
+ goto of_node_put;
+ }
+
ret = of_property_read_u32(port_np, "reg", &port_id);
if (ret < 0) {
dev_err(dev, "%pOF error reading port_id %d\n",
--
2.43.0



2024-05-03 23:25:04

by João Paulo Gonçalves

[permalink] [raw]
Subject: Re: [PATCH] net: ethernet: ti: am65-cpsw-nuss: create platform device for port nodes

On Fri, May 03, 2024 at 05:00:38PM -0300, Hiago De Franco wrote:
> From: Hiago De Franco <[email protected]>
>
> After this change, an 'of_node' link from '/sys/devices/platform' to
> '/sys/firmware/devicetree' will be created. The 'ethernet-ports' device
> allows multiple netdevs to have the exact same parent device, e.g. port@x
> netdevs are child nodes of ethernet-ports.
>
> When ethernet aliases are used (e.g. 'ethernet0 = &cpsw_port1' and
> 'ethernet1 = &cpsw_port2') in the device tree, without an of_node
> device exposed to the userspace, it is not possible to determine where
> exactly the alias is pointing to.
>
> As an example, this is essential for applications like systemd, which rely
> on the of_node information to identify and manage Ethernet devices
> using device tree aliases introduced in the v251 naming scheme.
>
> Signed-off-by: Hiago De Franco <[email protected]>
> ---
> drivers/net/ethernet/ti/am65-cpsw-nuss.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> index 1d00e21808c1..f74915f56fa2 100644
> --- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> +++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> @@ -2091,6 +2091,13 @@ static int am65_cpsw_nuss_init_slave_ports(struct am65_cpsw_common *common)
> if (strcmp(port_np->name, "port"))
> continue;
>
> + if (!of_platform_device_create(port_np, NULL, NULL)) {
> + dev_err(dev, "%pOF error creating port platform device\n",
> + port_np);
> + ret = -ENODEV;
> + goto of_node_put;
> + }
> +
> ret = of_property_read_u32(port_np, "reg", &port_id);
> if (ret < 0) {
> dev_err(dev, "%pOF error reading port_id %d\n",
> --
> 2.43.0
>
>

Reviewed-by: Joao Paulo Goncalves <[email protected]>

Regards,
João Paulo Gonçalves

2024-05-06 17:53:05

by Hiago De Franco

[permalink] [raw]
Subject: Re: [PATCH] net: ethernet: ti: am65-cpsw-nuss: create platform device for port nodes

Dear Maintainers,

On Fri, May 03, 2024 at 05:00:38PM -0300, Hiago De Franco wrote:
> From: Hiago De Franco <[email protected]>
>
> After this change, an 'of_node' link from '/sys/devices/platform' to
> '/sys/firmware/devicetree' will be created. The 'ethernet-ports' device
> allows multiple netdevs to have the exact same parent device, e.g. port@x
> netdevs are child nodes of ethernet-ports.
>
> When ethernet aliases are used (e.g. 'ethernet0 = &cpsw_port1' and
> 'ethernet1 = &cpsw_port2') in the device tree, without an of_node
> device exposed to the userspace, it is not possible to determine where
> exactly the alias is pointing to.
>
> As an example, this is essential for applications like systemd, which rely
> on the of_node information to identify and manage Ethernet devices
> using device tree aliases introduced in the v251 naming scheme.
>
> Signed-off-by: Hiago De Franco <[email protected]>
> ---
> drivers/net/ethernet/ti/am65-cpsw-nuss.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> index 1d00e21808c1..f74915f56fa2 100644
> --- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> +++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> @@ -2091,6 +2091,13 @@ static int am65_cpsw_nuss_init_slave_ports(struct am65_cpsw_common *common)
> if (strcmp(port_np->name, "port"))
> continue;
>
> + if (!of_platform_device_create(port_np, NULL, NULL)) {
> + dev_err(dev, "%pOF error creating port platform device\n",
> + port_np);
> + ret = -ENODEV;
> + goto of_node_put;
> + }
> +
> ret = of_property_read_u32(port_np, "reg", &port_id);
> if (ret < 0) {
> dev_err(dev, "%pOF error reading port_id %d\n",
> --
> 2.43.0
>

Apologies, I believe this patch is not 100% complete, there might be
something missing. Please keep this on hold, there is something not
working as expected that I am currently investigating.

My goal was to create an of_node device for the switch ports, which did
create under the /sys/devices/platform:

ls /sys/devices/platform/8000000.ethernet:ethernet-ports:port@1/
driver_override power uevent
modalias subsystem waiting_for_supplier
of_node supplier:platform:104044.phy

However it is still missing under

find /sys/devices/platform/bus@f0000/8000000.ethernet/net/eth0/ -name *of_node*

I will figure out how to do that and propose a new version.

Regards,

Hiago.