On Thu, 2017-01-19 at 13:22 +0100, Matthias Brugger wrote:

On 18/01/17 07:08, Chunfeng Yun wrote:

quoted

usually, the reference clock comes from 26M oscillator directly,
but some SoCs are not, add it for compatibility.

Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
---
 drivers/usb/mtu3/mtu3.h      |    1 +
 drivers/usb/mtu3/mtu3_plat.c |   21 +++++++++++++++++++--
 2 files changed, 20 insertions(+), 2 deletions(-)

[...]

quoted

@@ -154,6 +162,7 @@ static int ssusb_rscs_init(struct ssusb_mtk *ssusb)
 static void ssusb_rscs_exit(struct ssusb_mtk *ssusb)
 {
 	clk_disable_unprepare(ssusb->sys_clk);
+	clk_disable_unprepare(ssusb->ref_clk);
 	regulator_disable(ssusb->vusb33);
 	ssusb_phy_power_off(ssusb);
 	ssusb_phy_exit(ssusb);

@@ -216,6 +225,12 @@ static int get_ssusb_rscs(struct platform_device *pdev, struct ssusb_mtk *ssusb)
 		return PTR_ERR(ssusb->sys_clk);
 	}

+	ssusb->ref_clk = devm_clk_get(dev, "ref_ck");
+	if (IS_ERR(ssusb->ref_clk)) {
+		dev_err(dev, "failed to get ref clock\n");
+		return PTR_ERR(ssusb->ref_clk);
+	}
+

That would break older dts bindings, right?

Yes, So I send a new patch for the related dts. Maybe it's not a
problem, only one dts file need be updated currently.

ref_ck must be optional for the code.

I tend to make it be optional for the dts, but not for the code.
There are some "fixed-clock" which can be treated as dummy ones, and if
a clock is really optional, we can use one fixed-clock in dts, and keep
the code simple.
In fact, the reference clock is essential for usb controller.

Regards,
Matthias