On Wed, 2018-05-30 at 23:20 -0500, Bjorn Helgaas wrote:

On Wed, May 30, 2018 at 10:35:36AM +0800, honghui.zhang at mediatek.com wrote:

quoted

From: Honghui Zhang <redacted>

The MTCMOS of PCIe Host for MT2712 will be off when system suspend, and all
the internel control register will be reset after system resume. The PCIe
link should be re-established and the related control register values should
be re-set after system resume.

Signed-off-by: Honghui Zhang <redacted>
CC: Ryder Lee <ryder.lee@mediatek.com>
---
 drivers/pci/host/pcie-mediatek.c | 82 ++++++++++++++++++++++++++++++++++++++++
 1 file changed, 82 insertions(+)

diff --git a/drivers/pci/host/pcie-mediatek.c b/drivers/pci/host/pcie-mediatek.c
index dabf1086..60f98d92 100644
--- a/drivers/pci/host/pcie-mediatek.c
+++ b/drivers/pci/host/pcie-mediatek.c

@@ -132,12 +132,14 @@ struct mtk_pcie_port;
 /**
  * struct mtk_pcie_soc - differentiate between host generations
  * @need_fix_class_id: whether this host's class ID needed to be fixed or not
+ * @pm_support: whether the host's MTCMOS will be off when suspend
  * @ops: pointer to configuration access functions
  * @startup: pointer to controller setting functions
  * @setup_irq: pointer to initialize IRQ functions
  */
 struct mtk_pcie_soc {
 	bool need_fix_class_id;
+	bool pm_support;
 	struct pci_ops *ops;
 	int (*startup)(struct mtk_pcie_port *port);
 	int (*setup_irq)(struct mtk_pcie_port *port, struct device_node *node);

@@ -1179,12 +1181,91 @@ static int mtk_pcie_probe(struct platform_device *pdev)
 	return err;
 }
 
+static int __maybe_unused mtk_pcie_suspend_noirq(struct device *dev)
+{
+	struct platform_device *pdev;
+	struct mtk_pcie *pcie;
+	struct mtk_pcie_port *port;
+	const struct mtk_pcie_soc *soc;
+
+	pdev = to_platform_device(dev);
+	pcie = platform_get_drvdata(pdev);
+	soc = pcie->soc;
+	if (!soc->pm_support)
+		return 0;
+
+	list_for_each_entry(port, &pcie->ports, list) {
+		clk_disable_unprepare(port->ahb_ck);
+		clk_disable_unprepare(port->sys_ck);
+		phy_power_off(port->phy);
+	}
+
+	return 0;
+}
+
+static int __maybe_unused mtk_pcie_resume_noirq(struct device *dev)

I don't really like the __maybe_unused annotations.  Looking at the
other users of SET_NOIRQ_SYSTEM_SLEEP_PM_OPS, I think a small majority
of them wrap the function definitions in #ifdef CONFIG_PM_SLEEP
instead of using __maybe_unused.  That's also a bit ugly, but has the
advantage of truly omitting the definitions when they're not needed.

Ok, I will follow your advise in the next version.

thanks.