4.10-stable review patch. If anyone has any objections, please let me know. ------------------ From: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx> commit e7d5e412160c2143de1f818668774b33b3cdab0b upstream. Upon failing to acquire regulator supplies the qcom-ufs driver calls kfree() on the devm allocated memory used to store the name of the regulator, leading to devres corruption. Rather than switching to using the appropriate free function the patch acknowledge the fact that "name" is always a constant string and we don't actually need to create a local copy of it, but rather just reference the constant string. Fixes: add78fc05702 ("phy: qcom-ufs: Use devm sibling of kstrdup for regulator names") Reviewed-by: Subhash Jadavani <subhashj@xxxxxxxxxxxxxx> Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx> Signed-off-by: Kishon Vijay Abraham I <kishon@xxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- drivers/phy/phy-qcom-ufs.c | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) --- a/drivers/phy/phy-qcom-ufs.c +++ b/drivers/phy/phy-qcom-ufs.c @@ -217,12 +217,7 @@ static int __ufs_qcom_phy_init_vreg(stru char prop_name[MAX_PROP_NAME]; - vreg->name = devm_kstrdup(dev, name, GFP_KERNEL); - if (!vreg->name) { - err = -ENOMEM; - goto out; - } - + vreg->name = name; vreg->reg = devm_regulator_get(dev, name); if (IS_ERR(vreg->reg)) { err = PTR_ERR(vreg->reg); @@ -265,8 +260,6 @@ static int __ufs_qcom_phy_init_vreg(stru } out: - if (err) - kfree(vreg->name); return err; }