Hi,

On Sun, Mar 07, 2021 at 06:13:51AM +0300, Evgeny Boger wrote:
> R40 (aka V40/A40i/T3) and A10/A20 share the same EMAC IP.
> However, on R40 the EMAC is gated by default.
> 
> Signed-off-by: Evgeny Boger <boger@wirenboard.com>

On which device was it tested?

> ---
>  drivers/net/ethernet/allwinner/sun4i-emac.c | 21 ++++++++++++++++++++-
>  1 file changed, 20 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/allwinner/sun4i-emac.c b/drivers/net/ethernet/allwinner/sun4i-emac.c
> index 5ed80d9a6b9f..c0ae06dd922c 100644
> --- a/drivers/net/ethernet/allwinner/sun4i-emac.c
> +++ b/drivers/net/ethernet/allwinner/sun4i-emac.c
> @@ -28,6 +28,7 @@
>  #include <linux/of_platform.h>
>  #include <linux/platform_device.h>
>  #include <linux/phy.h>
> +#include <linux/reset.h>
>  #include <linux/soc/sunxi/sunxi_sram.h>
>  
>  #include "sun4i-emac.h"
> @@ -85,6 +86,7 @@ struct emac_board_info {
>  	unsigned int		link;
>  	unsigned int		speed;
>  	unsigned int		duplex;
> +	struct reset_control *reset;

You should align this with the rest of the other fields 

>  
>  	phy_interface_t		phy_interface;
>  };
> @@ -791,6 +793,7 @@ static int emac_probe(struct platform_device *pdev)
>  	struct net_device *ndev;
>  	int ret = 0;
>  	const char *mac_addr;
> +	struct reset_control *reset;
>  
>  	ndev = alloc_etherdev(sizeof(struct emac_board_info));
>  	if (!ndev) {
> @@ -852,6 +855,19 @@ static int emac_probe(struct platform_device *pdev)
>  		goto out_release_sram;
>  	}
>  
> +	reset = devm_reset_control_get_optional_exclusive(&pdev->dev, NULL);
> +	if (IS_ERR(reset)) {
> +		dev_err(&pdev->dev, "unable to request reset\n");
> +		ret = -ENODEV;
> +		goto out_release_sram;
> +	}

Judging from your commit log, it's not really optional for the R40. The
way we usually deal with this is to have a structure associated with a
new compatible and have a flag tell if that compatible requires a reset
line or not.

The dt binding should also be amended to allow the reset property

> +	db->reset = reset;
> +	ret = reset_control_deassert(db->reset);
> +	if (ret) {
> +		dev_err(&pdev->dev, "could not deassert EMAC reset\n");
> +		goto out_release_sram;
> +	}
> +

The programming guidelines in the datasheet ask that the reset line must
be deasserted before the clock in enabled.

Maxime