On Thu, Jul 17, 2025 at 03:23:38PM +0300, Laurentiu Palcu wrote:

Hi Frank,

On Wed, Jul 16, 2025 at 02:29:01PM -0400, Frank Li wrote:

quoted

On Wed, Jul 16, 2025 at 11:15:06AM +0300, Laurentiu Palcu wrote:

quoted

If runtime PM is used for the clock providers and they're part of a
power domain, then the power domain supply will be cut off when runtime
suspended. That means all BLK CTL registers belonging to that power
domain will be reset. Save the registers, then, before entering suspend
and restore them in resume.

Also, fix the suspend/resume routines and make sure we disable/enable
the clock correctly.

Signed-off-by: Laurentiu Palcu <laurentiu.palcu@oss.nxp.com>
---
 drivers/clk/imx/clk-imx95-blk-ctl.c | 25 +++++++++++++------------
 1 file changed, 13 insertions(+), 12 deletions(-)

diff --git a/drivers/clk/imx/clk-imx95-blk-ctl.c b/drivers/clk/imx/clk-imx95-blk-ctl.c
index c72debaf3a60b..3f6bcc33bbe99 100644
--- a/drivers/clk/imx/clk-imx95-blk-ctl.c
+++ b/drivers/clk/imx/clk-imx95-blk-ctl.c

@@ -453,7 +453,9 @@ static int imx95_bc_runtime_suspend(struct device *dev)
 {
 	struct imx95_blk_ctl *bc = dev_get_drvdata(dev);

+	bc->clk_reg_restore = readl(bc->base + bc->pdata->clk_reg_offset);
 	clk_disable_unprepare(bc->clk_apb);
+
 	return 0;
 }

@@ -461,7 +463,10 @@ static int imx95_bc_runtime_resume(struct device *dev)
 {
 	struct imx95_blk_ctl *bc = dev_get_drvdata(dev);

-	return clk_prepare_enable(bc->clk_apb);
+	clk_prepare_enable(bc->clk_apb);

Need check clk_prepare_enable()'s return value!

quoted

+	writel(bc->clk_reg_restore, bc->base + bc->pdata->clk_reg_offset);
+
+	return 0;
 }
 #endif

@@ -469,17 +474,12 @@ static int imx95_bc_runtime_resume(struct device *dev)
 static int imx95_bc_suspend(struct device *dev)
 {
 	struct imx95_blk_ctl *bc = dev_get_drvdata(dev);
-	int ret;

-	if (bc->pdata->rpm_enabled) {
-		ret = pm_runtime_get_sync(bc->dev);
-		if (ret < 0) {
-			pm_runtime_put_noidle(bc->dev);
-			return ret;
-		}
-	}
+	if (pm_runtime_suspended(dev))
+		return 0;

 	bc->clk_reg_restore = readl(bc->base + bc->pdata->clk_reg_offset);
+	clk_disable_unprepare(bc->clk_apb);

 	return 0;
 }

@@ -488,10 +488,11 @@ static int imx95_bc_resume(struct device *dev)
 {
 	struct imx95_blk_ctl *bc = dev_get_drvdata(dev);

-	writel(bc->clk_reg_restore, bc->base + bc->pdata->clk_reg_offset);
+	if (pm_runtime_suspended(dev))
+		return 0;

-	if (bc->pdata->rpm_enabled)
-		pm_runtime_put(bc->dev);
+	clk_prepare_enable(bc->clk_apb);
+	writel(bc->clk_reg_restore, bc->base + bc->pdata->clk_reg_offset);

 	return 0;
 }

Look like needn't imx95_bc_resume() and imx95_bc_suspend()

DEFINE_RUNTIME_DEV_PM_OPS() will use pm_runtime_force_suspend(), which
do similar things with above logic.

As I said for v1, we cannot use DEFINE_RUNTIME_DEV_PM_OPS(). This driver
is used for various clock providers and RPM can be disabled for some of
them (see rpm_enabled flag in platform data). When RPM is disabled and
DEFINE_RUNTIME_DEV_PM_OPS() is used, pm_runtime_force_suspend() is
called, as you pointed out, and suspend() is never called.

Sorry, I missed your message at v1. do you know which flag impact this?

Frank

Thanks,
Laurentiu

quoted

Frank

quoted

--
2.34.1