platform/x86/amd/pmf: Fix a double put in amd_pmf_remove()

[PATCH] platform/x86/amd/pmf: Fix a double put in amd_pmf_remove()

Posted by Christophe JAILLET 1 month, 2 weeks ago

The 'input_dev' is a managed resource allocated with
devm_input_allocate_device(), so there is no need to call
input_unregister_device() explicitly. It will be called automatically
when the driver is removed.

Fixes: 4c92d448e3e6 ("platform/x86/amd/pmf: Use existing input event codes to update system states")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
Compile tested-only

I'm not 100% confident with this change. The error handling scheme is not
a clear to me as what I usually see. For example, the last calls from
amd_pmf_probe() don't handle error at all. So the probe just succeeds in
these cases.

So, because of it, it is maybe fine to call input_unregister_device() in
amd_pmf_deinit_smart_pc(), even if it looks strange to me.

Review with care!
---
 drivers/platform/x86/amd/pmf/tee-if.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/platform/x86/amd/pmf/tee-if.c b/drivers/platform/x86/amd/pmf/tee-if.c
index e246367aacee..cc721fbc3e0b 100644
--- a/drivers/platform/x86/amd/pmf/tee-if.c
+++ b/drivers/platform/x86/amd/pmf/tee-if.c
@@ -515,9 +515,6 @@ int amd_pmf_init_smart_pc(struct amd_pmf_dev *dev)
 
 void amd_pmf_deinit_smart_pc(struct amd_pmf_dev *dev)
 {
-	if (dev->pmf_idev)
-		input_unregister_device(dev->pmf_idev);
-
 	if (pb_side_load && dev->esbin)
 		amd_pmf_remove_pb(dev);
 
-- 
2.45.2

Re: [PATCH] platform/x86/amd/pmf: Fix a double put in amd_pmf_remove()

Posted by Christophe JAILLET 1 month, 2 weeks ago

Le 25/07/2024 à 23:13, Christophe JAILLET a écrit :
> The 'input_dev' is a managed resource allocated with
> devm_input_allocate_device(), so there is no need to call
> input_unregister_device() explicitly. It will be called automatically
> when the driver is removed.
> 
> Fixes: 4c92d448e3e6 ("platform/x86/amd/pmf: Use existing input event codes to update system states")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
> Compile tested-only
> 
> I'm not 100% confident with this change. The error handling scheme is not
> a clear to me as what I usually see. For example, the last calls from
> amd_pmf_probe() don't handle error at all. So the probe just succeeds in
> these cases.
> 
> So, because of it, it is maybe fine to call input_unregister_device() in
> amd_pmf_deinit_smart_pc(), even if it looks strange to me.
> 
> Review with care!

NACK.

Things have been explained to me in another similar patch proposal.

Sorry for the noise.

CJ


> ---
>   drivers/platform/x86/amd/pmf/tee-if.c | 3 ---
>   1 file changed, 3 deletions(-)
> 
> diff --git a/drivers/platform/x86/amd/pmf/tee-if.c b/drivers/platform/x86/amd/pmf/tee-if.c
> index e246367aacee..cc721fbc3e0b 100644
> --- a/drivers/platform/x86/amd/pmf/tee-if.c
> +++ b/drivers/platform/x86/amd/pmf/tee-if.c
> @@ -515,9 +515,6 @@ int amd_pmf_init_smart_pc(struct amd_pmf_dev *dev)
>   
>   void amd_pmf_deinit_smart_pc(struct amd_pmf_dev *dev)
>   {
> -	if (dev->pmf_idev)
> -		input_unregister_device(dev->pmf_idev);
> -
>   	if (pb_side_load && dev->esbin)
>   		amd_pmf_remove_pb(dev);
>