Right now, we're blindly returning success even if waking the link fails during clock scaling. This is a mess because the core then tries to send SCSI commands to a dead link, causing huge timeouts.

Just return the actual error so the core can catch the failure and roll back properly.

Signed-off-by: Josh Law <objecting@objecting.org>
---
 drivers/ufs/host/ufs-qcom.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/ufs/host/ufs-qcom.c b/drivers/ufs/host/ufs-qcom.c
index 375fd24ba458..5d33a921a22f 100644
--- a/drivers/ufs/host/ufs-qcom.c
+++ b/drivers/ufs/host/ufs-qcom.c
@@ -1750,10 +1750,10 @@ static int ufs_qcom_clk_scale_notify(struct ufs_hba *hba, bool scale_up,
 		}
 
 		ufs_qcom_icc_update_bw(host);
-		ufshcd_uic_hibern8_exit(hba);
+		err = ufshcd_uic_hibern8_exit(hba);
 	}
 
-	return 0;
+	return err;
 }
 
 static void ufs_qcom_enable_test_bus(struct ufs_qcom_host *host)
-- 
2.34.1

On 29/03/2026 20:35, Josh Law wrote:
> Right now, we're blindly returning success even if waking the link fails during clock scaling. This is a mess because the core then tries to send SCSI commands to a dead link, causing huge timeouts.
> 
> Just return the actual error so the core can catch the failure and roll back properly.

NAK, you were asked way too many times for the same and you just ignore
people's feedback.

Best regards,
Krzysztof