[PATCH] phy: hisilicon: fix UTMI reset leaks in INNO USB2 PHY

[PATCH] phy: hisilicon: fix UTMI reset leaks in INNO USB2 PHY

[Xia Yan]

hisi_inno_phy_probe() acquires an exclusive UTMI reset control for each
child node but never releases it on probe failure or driver detach.
The leaked controls remain acquired, so a later probe can fail with
-EBUSY when requesting the same resets.

Register a devres action after each successful lookup to release the
control on probe failure and detach.

Found during source review of reset-control lifetimes.

Build-tested the driver object on x86_64 with W=1 and reset support
enabled. No hardware testing.

Fixes: ba8b0ee81fbb ("phy: add inno-usb2-phy driver for hi3798cv200 SoC")
Cc: stable@vger.kernel.org # needs of_node_put(child) on <= 6.6
Signed-off-by: Xia Yan <yanxia@cmss.chinamobile.com>
---
 drivers/phy/hisilicon/phy-hisi-inno-usb2.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/phy/hisilicon/phy-hisi-inno-usb2.c b/drivers/phy/hisilicon/phy-hisi-inno-usb2.c
index c843923252aa..9bf207e3e692 100644
--- a/drivers/phy/hisilicon/phy-hisi-inno-usb2.c
+++ b/drivers/phy/hisilicon/phy-hisi-inno-usb2.c
@@ -132,6 +132,11 @@ static const struct phy_ops hisi_inno_phy_ops = {
 	.owner = THIS_MODULE,
 };
 
+static void hisi_inno_phy_reset_put(void *data)
+{
+	reset_control_put(data);
+}
+
 static int hisi_inno_phy_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -169,6 +174,10 @@ static int hisi_inno_phy_probe(struct platform_device *pdev)
 		if (IS_ERR(rst))
 			return PTR_ERR(rst);
 
+		ret = devm_add_action_or_reset(dev, hisi_inno_phy_reset_put, rst);
+		if (ret)
+			return ret;
+
 		priv->ports[i].utmi_rst = rst;
 		priv->ports[i].priv = priv;
 

base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4

Re: [PATCH] phy: hisilicon: fix UTMI reset leaks in INNO USB2 PHY

[Philipp Zabel]

On Mi, 2026-11-11 at 16:59 +0800, Xia Yan wrote:
> hisi_inno_phy_probe() acquires an exclusive UTMI reset control for each
> child node but never releases it on probe failure or driver detach.
> The leaked controls remain acquired, so a later probe can fail with
> -EBUSY when requesting the same resets.
> 
> Register a devres action after each successful lookup to release the
> control on probe failure and detach.
> 
> Found during source review of reset-control lifetimes.
> 
> Build-tested the driver object on x86_64 with W=1 and reset support
> enabled. No hardware testing.
> 
> Fixes: ba8b0ee81fbb ("phy: add inno-usb2-phy driver for hi3798cv200 SoC")
> Cc: stable@vger.kernel.org # needs of_node_put(child) on <= 6.6
> Signed-off-by: Xia Yan <yanxia@cmss.chinamobile.com>

Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>

It looks like an equivalent to devm_get_clk_from_child() for reset
controls could be useful.

regards
Philipp