[PATCH] power: reset: msm: Process register_restart_handler() error

[PATCH] power: reset: msm: Process register_restart_handler() error

[Nikita Kiryushin]

If registering restart handler fails for msm-restart result is not checked.
It may be irrelevant now (as stated in comment to register_restart_handler,
the function currently always returns zero), but if the behavior changes
in the future, an error at registration of handler will be silently skipped.

Add return error code and print error message too debug log in case of
non-zero result of register_restart_handler.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 18a702e0de98 ("power: reset: use restart_notifier mechanism for 
msm-poweroff")

Signed-off-by: Nikita Kiryushin <kiryushin@ancud.ru>
---
  drivers/power/reset/msm-poweroff.c | 7 ++++++-
  1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/power/reset/msm-poweroff.c 
b/drivers/power/reset/msm-poweroff.c
index b9a401bd280b..5877a1ba2778 100644
--- a/drivers/power/reset/msm-poweroff.c
+++ b/drivers/power/reset/msm-poweroff.c
@@ -35,11 +35,16 @@ static void do_msm_poweroff(void)
   static int msm_restart_probe(struct platform_device *pdev)
  {
+	int ret = -EINVAL;
  	msm_ps_hold = devm_platform_ioremap_resource(pdev, 0);
  	if (IS_ERR(msm_ps_hold))
  		return PTR_ERR(msm_ps_hold);
  -	register_restart_handler(&restart_nb);
+	ret = register_restart_handler(&restart_nb);
+	if (ret) {
+		dev_err(&pdev->dev, "unable to register restart handler, %d\n", ret);
+		return ret;
+	}
   	pm_power_off = do_msm_poweroff;
  -- 2.34.1

Re: [PATCH] power: reset: msm: Process register_restart_handler() error

[Sebastian Reichel]

[-- Attachment #1: Type: text/plain, Size: 1929 bytes --]

Hi,

On Wed, Nov 08, 2023 at 08:27:57PM +0300, Nikita Kiryushin wrote:
> If registering restart handler fails for msm-restart result is not checked.
> It may be irrelevant now (as stated in comment to register_restart_handler,
> the function currently always returns zero), but if the behavior changes
> in the future, an error at registration of handler will be silently skipped.
> 
> Add return error code and print error message too debug log in case of
> non-zero result of register_restart_handler.
> 
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> 
> Fixes: 18a702e0de98 ("power: reset: use restart_notifier mechanism for
> msm-poweroff")
> 
> Signed-off-by: Nikita Kiryushin <kiryushin@ancud.ru>
> ---

If register_restart_handler fails, it might actually be a good idea
to continue and at least have a poweroff handler :)

>  drivers/power/reset/msm-poweroff.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/power/reset/msm-poweroff.c
> b/drivers/power/reset/msm-poweroff.c
> index b9a401bd280b..5877a1ba2778 100644
> --- a/drivers/power/reset/msm-poweroff.c
> +++ b/drivers/power/reset/msm-poweroff.c
> @@ -35,11 +35,16 @@ static void do_msm_poweroff(void)
>   static int msm_restart_probe(struct platform_device *pdev)
>  {
> +	int ret = -EINVAL;

no need to initialize, it's overwritten before being used.

>  	msm_ps_hold = devm_platform_ioremap_resource(pdev, 0);
>  	if (IS_ERR(msm_ps_hold))
>  		return PTR_ERR(msm_ps_hold);
>  -	register_restart_handler(&restart_nb);
> +	ret = register_restart_handler(&restart_nb);
> +	if (ret) {
> +		dev_err(&pdev->dev, "unable to register restart handler, %d\n", ret);
> +		return ret;

There is dev_err_probe() for this, but as mentioned above it's more
sensible not to fail here.

-- Sebastian

> +	}
>   	pm_power_off = do_msm_poweroff;
>  -- 2.34.1
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH] power: reset: msm: Process register_restart_handler() error

[Mukesh Ojha]

On 11/8/2023 10:57 PM, Nikita Kiryushin wrote:
> If registering restart handler fails for msm-restart result is not checked.
> It may be irrelevant now (as stated in comment to register_restart_handler,
> the function currently always returns zero), but if the behavior changes
> in the future, an error at registration of handler will be silently 
> skipped.
> 
> Add return error code and print error message too debug log in case of
> non-zero result of register_restart_handler.
> 
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
> 
> Fixes: 18a702e0de98 ("power: reset: use restart_notifier mechanism for 
> msm-poweroff")
> 
> Signed-off-by: Nikita Kiryushin <kiryushin@ancud.ru>
> ---
>   drivers/power/reset/msm-poweroff.c | 7 ++++++-
>   1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/power/reset/msm-poweroff.c 
> b/drivers/power/reset/msm-poweroff.c
> index b9a401bd280b..5877a1ba2778 100644
> --- a/drivers/power/reset/msm-poweroff.c
> +++ b/drivers/power/reset/msm-poweroff.c
> @@ -35,11 +35,16 @@ static void do_msm_poweroff(void)
>    static int msm_restart_probe(struct platform_device *pdev)
>   {
> +    int ret = -EINVAL;

This does not add up anything., no need to initialize.

-Mukesh


>       msm_ps_hold = devm_platform_ioremap_resource(pdev, 0);
>       if (IS_ERR(msm_ps_hold))
>           return PTR_ERR(msm_ps_hold);
>   -    register_restart_handler(&restart_nb);
> +    ret = register_restart_handler(&restart_nb);
> +    if (ret) {
> +        dev_err(&pdev->dev, "unable to register restart handler, %d\n", 
> ret);
> +        return ret;
> +    }
>        pm_power_off = do_msm_poweroff;
>   -- 2.34.1
> 

Re: [PATCH] power: reset: msm: Process register_restart_handler() error

[Guenter Roeck]

On 12/16/23 00:15, Mukesh Ojha wrote:
> 
> 
> On 11/8/2023 10:57 PM, Nikita Kiryushin wrote:
>> If registering restart handler fails for msm-restart result is not checked.
>> It may be irrelevant now (as stated in comment to register_restart_handler,
>> the function currently always returns zero), but if the behavior changes
>> in the future, an error at registration of handler will be silently skipped.
>>
>> Add return error code and print error message too debug log in case of
>> non-zero result of register_restart_handler.
>>
>> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>>
>> Fixes: 18a702e0de98 ("power: reset: use restart_notifier mechanism for msm-poweroff")
>>
>> Signed-off-by: Nikita Kiryushin <kiryushin@ancud.ru>
>> ---
>>   drivers/power/reset/msm-poweroff.c | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/power/reset/msm-poweroff.c b/drivers/power/reset/msm-poweroff.c
>> index b9a401bd280b..5877a1ba2778 100644
>> --- a/drivers/power/reset/msm-poweroff.c
>> +++ b/drivers/power/reset/msm-poweroff.c
>> @@ -35,11 +35,16 @@ static void do_msm_poweroff(void)
>>    static int msm_restart_probe(struct platform_device *pdev)
>>   {
>> +    int ret = -EINVAL;
> 
> This does not add up anything., no need to initialize.
> 
> -Mukesh
> 
> 
>>       msm_ps_hold = devm_platform_ioremap_resource(pdev, 0);
>>       if (IS_ERR(msm_ps_hold))
>>           return PTR_ERR(msm_ps_hold);
>>   -    register_restart_handler(&restart_nb);
>> +    ret = register_restart_handler(&restart_nb);

FWIW, checking this return value really do anything good either. The only possible
return value is -EEXIST, which would be returned if the restart handler was already
registered, and that can not really happen. On top of that, _if_ that happened,
the notifier core would dump a warning traceback, making the error message below
unnecessary.

>> +    if (ret) {
>> +        dev_err(&pdev->dev, "unable to register restart handler, %d\n", ret);
>> +        return ret;

On top of all the above, not registering the poweroff andler in the already
impossible case that the restart handler can not be registered seems like an
excessively bad idea to me.

Guenter

>> +    }
>>        pm_power_off = do_msm_poweroff;
>>   -- 2.34.1
>>