Re: [PATCH v2] net: stmmac: update Rx clk divider for 10M SGMII

[Andrew Halaney]

On Fri, Dec 01, 2023 at 03:35:48PM +0530, Sneh Shah wrote:
> SGMII 10MBPS mode needs RX clock divider to avoid drops in Rx.
> Update configure SGMII function with rx clk divider programming.
> 
> Fixes: 463120c31c58 ("net: stmmac: dwmac-qcom-ethqos: add support for SGMII")
> Signed-off-by: Sneh Shah <quic_snehshah@xxxxxxxxxxx>

Tested-by: Andrew Halaney <ahalaney@xxxxxxxxxx> # sa8775p-ride

I can confirm that without this patch traffic doesn't seem to work when
I force things in to SPEED_10 with ethtool (which required another
change to this driver in stmmac_ethtool.c to do so, which as far as I
understand Sarosh from your group will be upstreaming soon and is an
independent issue).

I am curious, I expected to see some dropped/errors output in ip link
when things weren't working. I guess the sgmii phy in the soc is
rejecting these prior to handing them to the mac? Or am I
misunderstanding how that works?

> ---
> v2 changelog:
> - Use FIELD_PREP to prepare bifield values in place of GENMASK
> - Add fixes tag
> ---
>  drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
> index d3bf42d0fceb..df6ff8bcdb5c 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
> @@ -34,6 +34,7 @@
>  #define RGMII_CONFIG_LOOPBACK_EN		BIT(2)
>  #define RGMII_CONFIG_PROG_SWAP			BIT(1)
>  #define RGMII_CONFIG_DDR_MODE			BIT(0)
> +#define RGMII_CONFIG_SGMII_CLK_DVDR		GENMASK(18, 10)
>  
>  /* SDCC_HC_REG_DLL_CONFIG fields */
>  #define SDCC_DLL_CONFIG_DLL_RST			BIT(30)
> @@ -617,6 +618,9 @@ static int ethqos_configure_sgmii(struct qcom_ethqos *ethqos)
>  	case SPEED_10:
>  		val |= ETHQOS_MAC_CTRL_PORT_SEL;
>  		val &= ~ETHQOS_MAC_CTRL_SPEED_MODE;
> +		rgmii_updatel(ethqos, RGMII_CONFIG_SGMII_CLK_DVDR,
> +			      FIELD_PREP(RGMII_CONFIG_SGMII_CLK_DVDR, 0x31),
> +			      RGMII_IO_MACRO_CONFIG);

Russell requested a comment about why you must program this every time,
I think it's a good idea too: https://lore.kernel.org/netdev/ZWch7LIqbMEaLRLW@xxxxxxxxxxxxxxxxxxxxx/

Also on my wishlist is making 0x31 less magic, but Santa might not bring
that for me :)

>  		break;
>  	}
>  
> -- 
> 2.17.1
> 
>