On 12/18/24 10:08, Manivannan Sadhasivam wrote:

On Tue, Dec 17, 2024 at 10:48:43AM +0100, Christian Bruel wrote:

quoted

On 12/16/24 17:17, Manivannan Sadhasivam wrote:

quoted

On Mon, Dec 16, 2024 at 11:02:07AM +0100, Christian Bruel wrote:

quoted

Hi Manivanna,

On 12/3/24 16:22, Manivannan Sadhasivam wrote:

quoted

On Tue, Nov 26, 2024 at 04:51:18PM +0100, Christian Bruel wrote:

[...]

quoted

+static int stm32_pcie_start_link(struct dw_pcie *pci)
+{
+	struct stm32_pcie *stm32_pcie = to_stm32_pcie(pci);
+	int ret;
+
+	if (stm32_pcie->link_status == STM32_PCIE_EP_LINK_ENABLED) {
+		dev_dbg(pci->dev, "Link is already enabled\n");
+		return 0;
+	}
+
+	ret = stm32_pcie_enable_link(pci);
+	if (ret) {
+		dev_err(pci->dev, "PCIe cannot establish link: %d\n", ret);
+		return ret;
+	}

How the REFCLK is supplied to the endpoint? From host or generated locally?

  From Host only, we don't support the separated clock model.

OK. So even without refclk you are still able to access the controller
registers? So the controller CSRs should be accessible by separate local clock I
believe.

Anyhow, please add this limitation (refclk dependency from host) in commit
message.

[...]

quoted

quoted

quoted

+	ret = phy_set_mode(stm32_pcie->phy, PHY_MODE_PCIE);

Hmm, so PHY mode is common for both endpoint and host?

Yes it is. We need to init the phy here because it is a clock source for the
PCIe core clk

Clock source? Is it coming directly to PCIe or through RCC? There is no direct
clock representation from PHY to PCIe in DT binding.

The core_clk is generated directly by the PLL in the COMBOPHY, gated by the
RCC

In that case, phy should be the clock provider to RCC and PCIe should get the
gated clock it.

ok, seems sensible.

- Mani