Re: [PATCH 2/2] PCI: mediatek-gen3: Add support for disable dvfsrc voltage request

[Jianjun Wang]

Hi Bjorn, Rob, Lorenzo,

Could you please help to take a look at this patch, I'm not sure if this
is a good solution, and I really need your suggestions.

Thanks.

On Fri, 2021-05-14 at 14:59 +0800, Jianjun Wang wrote:
> PCIe Gen3 PHY layer cannot work properly when the requested voltage
> is lower than a specific level(e.g. 0.55V, it's depends on
> the chip manufacturing process).
> 
> When the dvfsrc feature is implemented, the requested voltage
> may be reduced to a lower level in suspend mode, hence that
> the MAC layer will assert a HW signal to request the dvfsrc
> to raise voltage to normal mode, and it will wait the voltage
> ready signal which is derived from dvfsrc to determine whether
> the LTSSM can start normally.
> 
> When the dvfsrc feature is not implemented, the MAC layer still
> assert the voltage request to dvfsrc when exit suspend mode,
> but will not receive the voltage ready signal, in this case,
> the LTSSM cannot start normally, and the PCIe link will be failed.
> 
> Add support for disable dvfsrc voltage request. If the property of
> "disable-dvfsrc-vlt-req" is presented in device node, we assume that
> the requested voltage is always higher enough to keep the PCIe Gen3
> PHY active, and the voltage request to dvfsrc should be disabled.
> 
> Signed-off-by: Jianjun Wang <jianjun.wang@xxxxxxxxxxxx>
> ---
>  drivers/pci/controller/pcie-mediatek-gen3.c | 32 +++++++++++++++++++++
>  1 file changed, 32 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-mediatek-gen3.c b/drivers/pci/controller/pcie-mediatek-gen3.c
> index 20165e4a75b2..d1864303217e 100644
> --- a/drivers/pci/controller/pcie-mediatek-gen3.c
> +++ b/drivers/pci/controller/pcie-mediatek-gen3.c
> @@ -68,6 +68,9 @@
>  #define PCIE_MSI_SET_ENABLE_REG		0x190
>  #define PCIE_MSI_SET_ENABLE		GENMASK(PCIE_MSI_SET_NUM - 1, 0)
>  
> +#define PCIE_MISC_CTRL_REG		0x348
> +#define PCIE_DISABLE_DVFSRC_VLT_REQ	BIT(1)
> +
>  #define PCIE_MSI_SET_BASE_REG		0xc00
>  #define PCIE_MSI_SET_OFFSET		0x10
>  #define PCIE_MSI_SET_STATUS_OFFSET	0x04
> @@ -297,6 +300,35 @@ static int mtk_pcie_startup_port(struct mtk_pcie_port *port)
>  	val &= ~PCIE_INTX_ENABLE;
>  	writel_relaxed(val, port->base + PCIE_INT_ENABLE_REG);
>  
> +	/*
> +	 * PCIe Gen3 PHY layer can not work properly when the requested voltage
> +	 * is lower than a specific level(e.g. 0.55V, it's depends on
> +	 * the chip manufacturing process).
> +	 *
> +	 * When the dvfsrc feature is implemented, the requested voltage
> +	 * may be reduced to a lower level in suspend mode, hence that
> +	 * the MAC layer will assert a HW signal to request the dvfsrc
> +	 * to raise voltage to normal mode, and it will wait the voltage
> +	 * ready signal which is derived from dvfsrc to determine whether
> +	 * the LTSSM can start normally.
> +	 *
> +	 * When the dvfsrc feature is not implemented, the MAC layer still
> +	 * assert the voltage request to dvfsrc when exit suspend mode,
> +	 * but will not receive the voltage ready signal, in this case,
> +	 * the LTSSM cannot start normally, and the PCIe link will be failed.
> +	 *
> +	 * If the property of "disable-dvfsrc-vlt-req" is presented
> +	 * in device node, we assume that the requested voltage is always
> +	 * higher enough to keep the PCIe Gen3 PHY active, and the voltage
> +	 * request to dvfsrc should be disabled.
> +	 */
> +	val = readl_relaxed(port->base + PCIE_MISC_CTRL_REG);
> +	val &= ~PCIE_DISABLE_DVFSRC_VLT_REQ;
> +	if (of_property_read_bool(port->dev->of_node, "disable-dvfsrc-vlt-req"))
> +		val |= PCIE_DISABLE_DVFSRC_VLT_REQ;
> +
> +	writel(val, port->base + PCIE_MISC_CTRL_REG);
> +
>  	/* Assert all reset signals */
>  	val = readl_relaxed(port->base + PCIE_RST_CTRL_REG);
>  	val |= PCIE_MAC_RSTB | PCIE_PHY_RSTB | PCIE_BRG_RSTB | PCIE_PE_RSTB;