[PATCH] rpmsg: glink: Release driver_override

[PATCH] rpmsg: glink: Release driver_override

[Bjorn Andersson]

Upon termination of the rpmsg_device, driver_override needs to be freed
to avoid leaking the potentially assigned string.

Fixes: 42cd402b8fd4 ("rpmsg: Fix kfree() of static memory on setting driver_override")
Fixes: 39e47767ec9b ("rpmsg: Add driver_override device attribute for rpmsg_device")
Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
---
 drivers/rpmsg/qcom_glink_native.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/rpmsg/qcom_glink_native.c b/drivers/rpmsg/qcom_glink_native.c
index 8620eea1dc8a..046271e439c5 100644
--- a/drivers/rpmsg/qcom_glink_native.c
+++ b/drivers/rpmsg/qcom_glink_native.c
@@ -1454,6 +1454,7 @@ static void qcom_glink_rpdev_release(struct device *dev)
 {
 	struct rpmsg_device *rpdev = to_rpmsg_device(dev);
 
+	kfree(rpdev->driver_override);
 	kfree(rpdev);
 }
 
@@ -1697,6 +1698,7 @@ static void qcom_glink_device_release(struct device *dev)
 
 	/* Release qcom_glink_alloc_channel() reference */
 	kref_put(&channel->refcount, qcom_glink_channel_release);
+	kfree(rpdev->driver_override);
 	kfree(rpdev);
 }
 
-- 
2.37.3

Re: [PATCH] rpmsg: glink: Release driver_override

[Chris Lew]

On 1/9/2023 2:38 PM, Bjorn Andersson wrote:
> Upon termination of the rpmsg_device, driver_override needs to be freed
> to avoid leaking the potentially assigned string.
> 
> Fixes: 42cd402b8fd4 ("rpmsg: Fix kfree() of static memory on setting driver_override")
> Fixes: 39e47767ec9b ("rpmsg: Add driver_override device attribute for rpmsg_device")
> Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
> ---
>   drivers/rpmsg/qcom_glink_native.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/rpmsg/qcom_glink_native.c b/drivers/rpmsg/qcom_glink_native.c
> index 8620eea1dc8a..046271e439c5 100644
> --- a/drivers/rpmsg/qcom_glink_native.c
> +++ b/drivers/rpmsg/qcom_glink_native.c
> @@ -1454,6 +1454,7 @@ static void qcom_glink_rpdev_release(struct device *dev)
>   {
>   	struct rpmsg_device *rpdev = to_rpmsg_device(dev);
>   
> +	kfree(rpdev->driver_override);

I see in Krzysztof's original patch to fix up the driver_override we do 
kfree in other areas of rpmsg.

More of a just curious as to why don't we use the driver_set_override 
function to clear the string? It's mentioned in the rpmsg header that 
the helper should be used.

--- include/linux/rpmsg.h
  * @driver_override: driver name to force a match; do not set directly,
  *                   because core frees it; use driver_set_override() to
  *                   set or clear it.

>   	kfree(rpdev);
>   }
>   
> @@ -1697,6 +1698,7 @@ static void qcom_glink_device_release(struct device *dev)
>   
>   	/* Release qcom_glink_alloc_channel() reference */
>   	kref_put(&channel->refcount, qcom_glink_channel_release);
> +	kfree(rpdev->driver_override);
>   	kfree(rpdev);
>   }
>   
> 

Re: [PATCH] rpmsg: glink: Release driver_override

[Bjorn Andersson]

On Tue, Jan 17, 2023 at 05:01:01PM -0800, Chris Lew wrote:
> 
> 
> On 1/9/2023 2:38 PM, Bjorn Andersson wrote:
> > Upon termination of the rpmsg_device, driver_override needs to be freed
> > to avoid leaking the potentially assigned string.
> > 
> > Fixes: 42cd402b8fd4 ("rpmsg: Fix kfree() of static memory on setting driver_override")
> > Fixes: 39e47767ec9b ("rpmsg: Add driver_override device attribute for rpmsg_device")
> > Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
> > ---
> >   drivers/rpmsg/qcom_glink_native.c | 2 ++
> >   1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/rpmsg/qcom_glink_native.c b/drivers/rpmsg/qcom_glink_native.c
> > index 8620eea1dc8a..046271e439c5 100644
> > --- a/drivers/rpmsg/qcom_glink_native.c
> > +++ b/drivers/rpmsg/qcom_glink_native.c
> > @@ -1454,6 +1454,7 @@ static void qcom_glink_rpdev_release(struct device *dev)
> >   {
> >   	struct rpmsg_device *rpdev = to_rpmsg_device(dev);
> > +	kfree(rpdev->driver_override);
> 
> I see in Krzysztof's original patch to fix up the driver_override we do
> kfree in other areas of rpmsg.
> 

Right, in the original patch what became driver_set_override() was open
coded and hence there was a kfree().

> More of a just curious as to why don't we use the driver_set_override
> function to clear the string? It's mentioned in the rpmsg header that the
> helper should be used.
> 
> --- include/linux/rpmsg.h
>  * @driver_override: driver name to force a match; do not set directly,
>  *                   because core frees it; use driver_set_override() to
>  *                   set or clear it.
> 

Looking around the kernel, just calling kfree() in the release function
seems to be how others are doing it as well. As such I presume that
comment applies to the runtime state, rather than while we're cleaning
things up.

Regards,
Bjorn

> >   	kfree(rpdev);
> >   }
> > @@ -1697,6 +1698,7 @@ static void qcom_glink_device_release(struct device *dev)
> >   	/* Release qcom_glink_alloc_channel() reference */
> >   	kref_put(&channel->refcount, qcom_glink_channel_release);
> > +	kfree(rpdev->driver_override);
> >   	kfree(rpdev);
> >   }
> > 

Re: [PATCH] rpmsg: glink: Release driver_override

[Chris Lew]

On 1/18/2023 9:50 AM, Bjorn Andersson wrote:
> On Tue, Jan 17, 2023 at 05:01:01PM -0800, Chris Lew wrote:
>>
>>
>> On 1/9/2023 2:38 PM, Bjorn Andersson wrote:
>>> Upon termination of the rpmsg_device, driver_override needs to be freed
>>> to avoid leaking the potentially assigned string.
>>>
>>> Fixes: 42cd402b8fd4 ("rpmsg: Fix kfree() of static memory on setting driver_override")
>>> Fixes: 39e47767ec9b ("rpmsg: Add driver_override device attribute for rpmsg_device")
>>> Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
>>> ---
>>>    drivers/rpmsg/qcom_glink_native.c | 2 ++
>>>    1 file changed, 2 insertions(+)
>>>
>>> diff --git a/drivers/rpmsg/qcom_glink_native.c b/drivers/rpmsg/qcom_glink_native.c
>>> index 8620eea1dc8a..046271e439c5 100644
>>> --- a/drivers/rpmsg/qcom_glink_native.c
>>> +++ b/drivers/rpmsg/qcom_glink_native.c
>>> @@ -1454,6 +1454,7 @@ static void qcom_glink_rpdev_release(struct device *dev)
>>>    {
>>>    	struct rpmsg_device *rpdev = to_rpmsg_device(dev);
>>> +	kfree(rpdev->driver_override);
>>
>> I see in Krzysztof's original patch to fix up the driver_override we do
>> kfree in other areas of rpmsg.
>>
> 
> Right, in the original patch what became driver_set_override() was open
> coded and hence there was a kfree().
> 
>> More of a just curious as to why don't we use the driver_set_override
>> function to clear the string? It's mentioned in the rpmsg header that the
>> helper should be used.
>>
>> --- include/linux/rpmsg.h
>>   * @driver_override: driver name to force a match; do not set directly,
>>   *                   because core frees it; use driver_set_override() to
>>   *                   set or clear it.
>>
> 
> Looking around the kernel, just calling kfree() in the release function
> seems to be how others are doing it as well. As such I presume that
> comment applies to the runtime state, rather than while we're cleaning
> things up.
> 
> Regards,
> Bjorn
> 

Got it - should be safe to follow how others are handling this in cleanup.

Reviewed-by: Chris Lew <quic_clew@quicinc.com>

>>>    	kfree(rpdev);
>>>    }
>>> @@ -1697,6 +1698,7 @@ static void qcom_glink_device_release(struct device *dev)
>>>    	/* Release qcom_glink_alloc_channel() reference */
>>>    	kref_put(&channel->refcount, qcom_glink_channel_release);
>>> +	kfree(rpdev->driver_override);
>>>    	kfree(rpdev);
>>>    }
>>>