ufs_qcom_enable_lane_clks() and ufs_qcom_disable_lane_clks() currently
use clk_bulk_prepare_enable()/clk_bulk_disable_unprepare() on the
entire host->clks array obtained from devm_clk_bulk_get_all(). This
array contains all device clocks, not just lane symbol clocks.

Since the UFS core framework already manages the non-lane clocks via
the setup_clocks callback, the bulk enable/disable in the lane clock
APIs resulted in duplicate reference count increments on those shared
clocks. The extra enable counts were never balanced by a corresponding
disable from the framework's clock gating path, preventing the clock
reference counts from reaching zero and ultimately blocking CXO
shutdown during low-power states.

Fix this by restricting the lane clock APIs to only prepare/enable
and disable/unprepare the three lane symbol clocks (tx_lane0_sync_clk,
rx_lane0_sync_clk, rx_lane1_sync_clk), leaving the handling of all
other clocks to the UFS core framewore.

Signed-off-by: Nitin Rawat <nitin.rawat@oss.qualcomm.com>
---
 drivers/ufs/host/ufs-qcom.c | 41 ++++++++++++++++++++++++++++++++-----
 drivers/ufs/host/ufs-qcom.h |  3 +++
 2 files changed, 39 insertions(+), 5 deletions(-)

diff --git a/drivers/ufs/host/ufs-qcom.c b/drivers/ufs/host/ufs-qcom.c
index 291c43448764..7ec1626704c6 100644
--- a/drivers/ufs/host/ufs-qcom.c
+++ b/drivers/ufs/host/ufs-qcom.c
@@ -347,7 +347,9 @@ static void ufs_qcom_disable_lane_clks(struct ufs_qcom_host *host)
 	if (!host->is_lane_clks_enabled)
 		return;

-	clk_bulk_disable_unprepare(host->num_clks, host->clks);
+	clk_disable_unprepare(host->rx_lane1_sync_clk);
+	clk_disable_unprepare(host->rx_lane0_sync_clk);
+	clk_disable_unprepare(host->tx_lane0_sync_clk);

 	host->is_lane_clks_enabled = false;
 }
@@ -356,18 +358,35 @@ static int ufs_qcom_enable_lane_clks(struct ufs_qcom_host *host)
 {
 	int err;

-	err = clk_bulk_prepare_enable(host->num_clks, host->clks);
+	if (host->is_lane_clks_enabled)
+		return 0;
+
+	err = clk_prepare_enable(host->tx_lane0_sync_clk);
 	if (err)
-		return err;
+		goto out;

-	host->is_lane_clks_enabled = true;
+	err = clk_prepare_enable(host->rx_lane0_sync_clk);
+	if (err)
+		goto out_disable_tx_lane0;
+
+	err = clk_prepare_enable(host->rx_lane1_sync_clk);
+	if (err)
+		goto out_disable_rx_lane0;

+	host->is_lane_clks_enabled = true;
 	return 0;
+
+out_disable_rx_lane0:
+	clk_disable_unprepare(host->rx_lane0_sync_clk);
+out_disable_tx_lane0:
+	clk_disable_unprepare(host->tx_lane0_sync_clk);
+out:
+	return err;
 }

 static int ufs_qcom_init_lane_clks(struct ufs_qcom_host *host)
 {
-	int err;
+	int err, i;
 	struct device *dev = host->hba->dev;

 	if (has_acpi_companion(dev))
@@ -379,6 +398,18 @@ static int ufs_qcom_init_lane_clks(struct ufs_qcom_host *host)

 	host->num_clks = err;

+	for (i = 0; i < host->num_clks; i++) {
+		if (!host->clks[i].id)
+			continue;
+		if (!strcmp(host->clks[i].id, "tx_lane0_sync_clk"))
+			host->tx_lane0_sync_clk = host->clks[i].clk;
+		else if (!strcmp(host->clks[i].id, "rx_lane0_sync_clk"))
+			host->rx_lane0_sync_clk = host->clks[i].clk;
+		else if (!strcmp(host->clks[i].id, "rx_lane1_sync_clk"))
+			if (host->hba->lanes_per_direction > 1)
+				host->rx_lane1_sync_clk = host->clks[i].clk;
+	}
+
 	return 0;
 }

diff --git a/drivers/ufs/host/ufs-qcom.h b/drivers/ufs/host/ufs-qcom.h
index e20b3ca50577..2fcc3bc1133d 100644
--- a/drivers/ufs/host/ufs-qcom.h
+++ b/drivers/ufs/host/ufs-qcom.h
@@ -331,6 +331,9 @@ struct ufs_qcom_host {
 	struct ufs_hba *hba;
 	struct ufs_pa_layer_attr dev_req_params;
 	struct clk_bulk_data *clks;
+	struct clk *tx_lane0_sync_clk;
+	struct clk *rx_lane0_sync_clk;
+	struct clk *rx_lane1_sync_clk;
 	u32 num_clks;
 	bool is_lane_clks_enabled;

--
2.34.1

On 7/13/26 7:34 PM, Nitin Rawat wrote:
> ufs_qcom_enable_lane_clks() and ufs_qcom_disable_lane_clks() currently
> use clk_bulk_prepare_enable()/clk_bulk_disable_unprepare() on the
> entire host->clks array obtained from devm_clk_bulk_get_all(). This
> array contains all device clocks, not just lane symbol clocks.
> 
> Since the UFS core framework already manages the non-lane clocks via
> the setup_clocks callback, the bulk enable/disable in the lane clock
> APIs resulted in duplicate reference count increments on those shared
> clocks. The extra enable counts were never balanced by a corresponding
> disable from the framework's clock gating path, preventing the clock
> reference counts from reaching zero and ultimately blocking CXO
> shutdown during low-power states.
> 
> Fix this by restricting the lane clock APIs to only prepare/enable
> and disable/unprepare the three lane symbol clocks (tx_lane0_sync_clk,
> rx_lane0_sync_clk, rx_lane1_sync_clk), leaving the handling of all
> other clocks to the UFS core framewore.
> 
> Signed-off-by: Nitin Rawat <nitin.rawat@oss.qualcomm.com>
> ---
>  drivers/ufs/host/ufs-qcom.c | 41 ++++++++++++++++++++++++++++++++-----
>  drivers/ufs/host/ufs-qcom.h |  3 +++
>  2 files changed, 39 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/ufs/host/ufs-qcom.c b/drivers/ufs/host/ufs-qcom.c
> index 291c43448764..7ec1626704c6 100644
> --- a/drivers/ufs/host/ufs-qcom.c
> +++ b/drivers/ufs/host/ufs-qcom.c
> @@ -347,7 +347,9 @@ static void ufs_qcom_disable_lane_clks(struct ufs_qcom_host *host)
>  	if (!host->is_lane_clks_enabled)
>  		return;
> 
> -	clk_bulk_disable_unprepare(host->num_clks, host->clks);
> +	clk_disable_unprepare(host->rx_lane1_sync_clk);
> +	clk_disable_unprepare(host->rx_lane0_sync_clk);
> +	clk_disable_unprepare(host->tx_lane0_sync_clk);

I don't know if this matters (probably not since IIUC the controller
would be suspended already), but downstream disables them in the
following order:

TX_L1
TX_L0
RX_L1
RX_L0

> 
>  	host->is_lane_clks_enabled = false;
>  }
> @@ -356,18 +358,35 @@ static int ufs_qcom_enable_lane_clks(struct ufs_qcom_host *host)
>  {
>  	int err;
> 
> -	err = clk_bulk_prepare_enable(host->num_clks, host->clks);
> +	if (host->is_lane_clks_enabled)
> +		return 0;
> +
> +	err = clk_prepare_enable(host->tx_lane0_sync_clk);
>  	if (err)
> -		return err;
> +		goto out;
> 
> -	host->is_lane_clks_enabled = true;
> +	err = clk_prepare_enable(host->rx_lane0_sync_clk);
> +	if (err)
> +		goto out_disable_tx_lane0;
> +
> +	err = clk_prepare_enable(host->rx_lane1_sync_clk);
> +	if (err)
> +		goto out_disable_rx_lane0;
> 
> +	host->is_lane_clks_enabled = true;
>  	return 0;

Let's keep the \n above return

> +
> +out_disable_rx_lane0:
> +	clk_disable_unprepare(host->rx_lane0_sync_clk);
> +out_disable_tx_lane0:
> +	clk_disable_unprepare(host->tx_lane0_sync_clk);
> +out:
> +	return err;
>  }
> 
>  static int ufs_qcom_init_lane_clks(struct ufs_qcom_host *host)
>  {
> -	int err;
> +	int err, i;
>  	struct device *dev = host->hba->dev;
> 
>  	if (has_acpi_companion(dev))
> @@ -379,6 +398,18 @@ static int ufs_qcom_init_lane_clks(struct ufs_qcom_host *host)
> 
>  	host->num_clks = err;
> 
> +	for (i = 0; i < host->num_clks; i++) {
> +		if (!host->clks[i].id)
> +			continue;
> +		if (!strcmp(host->clks[i].id, "tx_lane0_sync_clk"))
> +			host->tx_lane0_sync_clk = host->clks[i].clk;
> +		else if (!strcmp(host->clks[i].id, "rx_lane0_sync_clk"))
> +			host->rx_lane0_sync_clk = host->clks[i].clk;
> +		else if (!strcmp(host->clks[i].id, "rx_lane1_sync_clk"))
> +			if (host->hba->lanes_per_direction > 1)
> +				host->rx_lane1_sync_clk = host->clks[i].clk;
> +	}

This is technically probably a little faster than calling
clk_get() again, but please do that instead

Konrad