On Wed, Aug 20, 2025 at 03:35:36PM -0500, Bjorn Helgaas wrote:

On Wed, Aug 20, 2025 at 04:10:48PM +0800, Richard Zhu wrote:

quoted

The CLKREQ# is an open drain, active low signal that is driven low by
the card to request reference clock.

Since the reference clock may be required by i.MX PCIe host too. To make
sure this clock is available even when the CLKREQ# isn't driven low by
the card(e.x no card connected), force CLKREQ# override active low for
i.MX PCIe host during initialization.

The CLKREQ# override can be cleared safely when supports-clkreq is
present and PCIe link is up later. Because the CLKREQ# would be driven
low by the card in this case.

Signed-off-by: Richard Zhu <hongxing.zhu@nxp.com>
---
 drivers/pci/controller/dwc/pci-imx6.c | 35 +++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

diff --git a/drivers/pci/controller/dwc/pci-imx6.c b/drivers/pci/controller/dwc/pci-imx6.c
index 80e48746bbaf6..a73632b47e2d3 100644
--- a/drivers/pci/controller/dwc/pci-imx6.c
+++ b/drivers/pci/controller/dwc/pci-imx6.c

@@ -52,6 +52,8 @@
 #define IMX95_PCIE_REF_CLKEN			BIT(23)
 #define IMX95_PCIE_PHY_CR_PARA_SEL		BIT(9)
 #define IMX95_PCIE_SS_RW_REG_1			0xf4
+#define IMX95_PCIE_CLKREQ_OVERRIDE_EN		BIT(8)
+#define IMX95_PCIE_CLKREQ_OVERRIDE_VAL		BIT(9)
 #define IMX95_PCIE_SYS_AUX_PWR_DET		BIT(31)

 #define IMX95_PE0_GEN_CTRL_1			0x1050

@@ -136,6 +138,7 @@ struct imx_pcie_drvdata {
 	int (*enable_ref_clk)(struct imx_pcie *pcie, bool enable);
 	int (*core_reset)(struct imx_pcie *pcie, bool assert);
 	int (*wait_pll_lock)(struct imx_pcie *pcie);
+	void (*clr_clkreq_override)(struct imx_pcie *pcie);
 	const struct dw_pcie_host_ops *ops;
 };

@@ -149,6 +152,7 @@ struct imx_pcie {
 	struct gpio_desc	*reset_gpiod;
 	struct clk_bulk_data	*clks;
 	int			num_clks;
+	bool			supports_clkreq;
 	struct regmap		*iomuxc_gpr;
 	u16			msi_ctrl;
 	u32			controller_id;

@@ -267,6 +271,13 @@ static int imx95_pcie_init_phy(struct imx_pcie *imx_pcie)
 			   IMX95_PCIE_REF_CLKEN,
 			   IMX95_PCIE_REF_CLKEN);

+	/* Force CLKREQ# low by override */
+	regmap_update_bits(imx_pcie->iomuxc_gpr,
+			   IMX95_PCIE_SS_RW_REG_1,
+			   IMX95_PCIE_CLKREQ_OVERRIDE_EN |
+			   IMX95_PCIE_CLKREQ_OVERRIDE_VAL,
+			   IMX95_PCIE_CLKREQ_OVERRIDE_EN |
+			   IMX95_PCIE_CLKREQ_OVERRIDE_VAL);
 	return 0;
 }

@@ -1298,6 +1309,18 @@ static void imx_pcie_host_exit(struct dw_pcie_rp *pp)
 		regulator_disable(imx_pcie->vpcie);
 }

+static void imx8mm_pcie_clr_clkreq_override(struct imx_pcie *imx_pcie)
+{
+	imx8mm_pcie_enable_ref_clk(imx_pcie, false);
+}
+
+static void imx95_pcie_clr_clkreq_override(struct imx_pcie *imx_pcie)
+{
+	regmap_update_bits(imx_pcie->iomuxc_gpr, IMX95_PCIE_SS_RW_REG_1,
+			   IMX95_PCIE_CLKREQ_OVERRIDE_EN |
+			   IMX95_PCIE_CLKREQ_OVERRIDE_VAL, 0);
+}
+
 static void imx_pcie_host_post_init(struct dw_pcie_rp *pp)
 {
 	struct dw_pcie *pci = to_dw_pcie_from_pp(pp);

@@ -1322,6 +1345,12 @@ static void imx_pcie_host_post_init(struct dw_pcie_rp *pp)
 		dw_pcie_writel_dbi(pci, GEN3_RELATED_OFF, val);
 		dw_pcie_dbi_ro_wr_dis(pci);
 	}
+
+	/* Clear CLKREQ# override if supports_clkreq is true and link is up */
+	if (dw_pcie_link_up(pci) && imx_pcie->supports_clkreq) {
+		if (imx_pcie->drvdata->clr_clkreq_override)
+			imx_pcie->drvdata->clr_clkreq_override(imx_pcie);

It seems racy to clear the override when the link is up.

Since it sounds like the i.MX host requires refclock all the time, why
not keep the override permanently?

It will Lost l1ss support if enable override permanetly. I think other
platform have similar issues (at least dwc controller). Most platform use
m.2 socket, which pull down clk_req by EP side devices.

Only standard PCIe slots,  which clkreq have not connect by EP devices
because stardard PCIe slots add #clkreq signal happen recent years (maybe
after miniPCIe slot spec). Some old PCIe devices have not connect it.

Even latest PCIe devices, I saw have jumper in PCIe card to controller
#clkreq.

Obviously it must be ok to keep the override for a while, because
there is some interval between the link coming up and the call of
.clr_clkreq_override().

Would something bad happen if we *never* called
.clr_clkreq_override()?

clock will always running, lost L1SS power save feature.

Frank

Bjorn