| Message ID | 20241030-cpuidle-qcom-spm-cleanup-v1-2-04416fcca7de@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | cpuidle: qcom-spm: fix resource release in spm_cpuidle_register | expand |
On Wed, Oct 30, 2024 at 07:38:33AM +0100, Javier Carrasco wrote: > A reference to a device obtained via of_find_device_by_node() requires > explicit calls to put_device() when it is no longer required to avoid > leaking the resource. > > Add the missing calls to put_device(&pdev->dev) in the success path as > well as in the only error path before the device is no longer required. > > Note that the acquired device is neither assigned nor used to manage > additional resources, and it can be released right after using it. Well... This raises one question: if the device is put, then its drvdata can go away. But at the same time the drvdata can also go away if the SPM device is just unbound from the driver. Granted the age of those platforms it's probably not worth refactoring the drivers too much. Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org> > > Cc: stable@vger.kernel.org > Fixes: 60f3692b5f0b ("cpuidle: qcom_spm: Detach state machine from main SPM handling") > Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com> > --- > drivers/cpuidle/cpuidle-qcom-spm.c | 5 ++++- > 1 file changed, 4 insertions(+), 1 deletion(-) > > diff --git a/drivers/cpuidle/cpuidle-qcom-spm.c b/drivers/cpuidle/cpuidle-qcom-spm.c > index c9ab49b310fd..601aa81ffff3 100644 > --- a/drivers/cpuidle/cpuidle-qcom-spm.c > +++ b/drivers/cpuidle/cpuidle-qcom-spm.c > @@ -106,10 +106,13 @@ static int spm_cpuidle_register(struct device *cpuidle_dev, int cpu) > return -ENODEV; > > data = devm_kzalloc(cpuidle_dev, sizeof(*data), GFP_KERNEL); > - if (!data) > + if (!data) { > + put_device(&pdev->dev); > return -ENOMEM; > + } > > data->spm = dev_get_drvdata(&pdev->dev); > + put_device(&pdev->dev); > if (!data->spm) > return -EINVAL; > > > -- > 2.43.0 >
diff --git a/drivers/cpuidle/cpuidle-qcom-spm.c b/drivers/cpuidle/cpuidle-qcom-spm.c index c9ab49b310fd..601aa81ffff3 100644 --- a/drivers/cpuidle/cpuidle-qcom-spm.c +++ b/drivers/cpuidle/cpuidle-qcom-spm.c @@ -106,10 +106,13 @@ static int spm_cpuidle_register(struct device *cpuidle_dev, int cpu) return -ENODEV; data = devm_kzalloc(cpuidle_dev, sizeof(*data), GFP_KERNEL); - if (!data) + if (!data) { + put_device(&pdev->dev); return -ENOMEM; + } data->spm = dev_get_drvdata(&pdev->dev); + put_device(&pdev->dev); if (!data->spm) return -EINVAL;
A reference to a device obtained via of_find_device_by_node() requires explicit calls to put_device() when it is no longer required to avoid leaking the resource. Add the missing calls to put_device(&pdev->dev) in the success path as well as in the only error path before the device is no longer required. Note that the acquired device is neither assigned nor used to manage additional resources, and it can be released right after using it. Cc: stable@vger.kernel.org Fixes: 60f3692b5f0b ("cpuidle: qcom_spm: Detach state machine from main SPM handling") Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com> --- drivers/cpuidle/cpuidle-qcom-spm.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)