[PATCH 2/2] ARM: davinci: PM: Do not free useful resources in normal path in 'davinci_pm_init'

[PATCH 2/2] ARM: davinci: PM: Do not free useful resources in normal path in 'davinci_pm_init'

[Christophe JAILLET]

This looks spurious to iounmap resources in the normal path of this init
function.
The 3 ioremap'ed fields of 'pm_config' can be accessed later on in other
functions, so it is likely that we should return 'success' before unrolling
everything.

Fixes: aa9aa1ec2df6 ("ARM: davinci: PM: rework init, remove platform device")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
This patch is just a *guess*. The end of the function looks more like a
error handling code rather than a normal path.
---
 arch/arm/mach-davinci/pm.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm/mach-davinci/pm.c b/arch/arm/mach-davinci/pm.c
index d282b0783ecf..163d865abbf9 100644
--- a/arch/arm/mach-davinci/pm.c
+++ b/arch/arm/mach-davinci/pm.c
@@ -161,6 +161,7 @@ int __init davinci_pm_init(void)
 						davinci_cpu_suspend_sz);
 
 	suspend_set_ops(&davinci_pm_ops);
+	return 0;
 
 no_sram_mem:
 	iounmap(pm_config.ddrpsc_reg_base);
-- 
2.11.0

Re: [PATCH 2/2] ARM: davinci: PM: Do not free useful resources in normal path in 'davinci_pm_init'

[walter harms]

Am 13.05.2017 13:40, schrieb Christophe JAILLET:
> This looks spurious to iounmap resources in the normal path of this init
> function.
> The 3 ioremap'ed fields of 'pm_config' can be accessed later on in other
> functions, so it is likely that we should return 'success' before unrolling
> everything.
> 
> Fixes: aa9aa1ec2df6 ("ARM: davinci: PM: rework init, remove platform device")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
> This patch is just a *guess*. The end of the function looks more like a
> error handling code rather than a normal path.
> ---
>  arch/arm/mach-davinci/pm.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/arm/mach-davinci/pm.c b/arch/arm/mach-davinci/pm.c
> index d282b0783ecf..163d865abbf9 100644
> --- a/arch/arm/mach-davinci/pm.c
> +++ b/arch/arm/mach-davinci/pm.c
> @@ -161,6 +161,7 @@ int __init davinci_pm_init(void)
>  						davinci_cpu_suspend_sz);
>  
>  	suspend_set_ops(&davinci_pm_ops);
> +	return 0;
>  
>  no_sram_mem:
>  	iounmap(pm_config.ddrpsc_reg_base);


looks like, but that would mean that is wrong also:

	davinci_sram_suspend = sram_alloc(davinci_cpu_suspend_sz, NULL);
	if (!davinci_sram_suspend) {
		pr_err("PM: cannot allocate SRAM memory\n");
		return -ENOMEM;
	}

what means 1 iounmap missing.

re,
 wh

Re: [PATCH 2/2] ARM: davinci: PM: Do not free useful resources in normal path in 'davinci_pm_init'

[Christophe JAILLET]

Le 13/05/2017 à 15:22, walter harms a écrit :
>
> Am 13.05.2017 13:40, schrieb Christophe JAILLET:
>> This looks spurious to iounmap resources in the normal path of this init
>> function.
>> The 3 ioremap'ed fields of 'pm_config' can be accessed later on in other
>> functions, so it is likely that we should return 'success' before unrolling
>> everything.
>>
>> Fixes: aa9aa1ec2df6 ("ARM: davinci: PM: rework init, remove platform device")
>> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
>> ---
>> This patch is just a *guess*. The end of the function looks more like a
>> error handling code rather than a normal path.
>> ---
>>   arch/arm/mach-davinci/pm.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/arch/arm/mach-davinci/pm.c b/arch/arm/mach-davinci/pm.c
>> index d282b0783ecf..163d865abbf9 100644
>> --- a/arch/arm/mach-davinci/pm.c
>> +++ b/arch/arm/mach-davinci/pm.c
>> @@ -161,6 +161,7 @@ int __init davinci_pm_init(void)
>>   						davinci_cpu_suspend_sz);
>>   
>>   	suspend_set_ops(&davinci_pm_ops);
>> +	return 0;
>>   
>>   no_sram_mem:
>>   	iounmap(pm_config.ddrpsc_reg_base);
>
> looks like, but that would mean that is wrong also:
>
> 	davinci_sram_suspend = sram_alloc(davinci_cpu_suspend_sz, NULL);
> 	if (!davinci_sram_suspend) {
> 		pr_err("PM: cannot allocate SRAM memory\n");
> 		return -ENOMEM;
> 	}
>
> what means 1 iounmap missing.
>
> re,
>   wh
>
This is what I try to fix in the [1/2] patch.

CJ

Re: [PATCH 2/2] ARM: davinci: PM: Do not free useful resources in normal path in 'davinci_pm_init'

[Sekhar Nori]

On Saturday 13 May 2017 05:10 PM, Christophe JAILLET wrote:
> This looks spurious to iounmap resources in the normal path of this init
> function.
> The 3 ioremap'ed fields of 'pm_config' can be accessed later on in other
> functions, so it is likely that we should return 'success' before unrolling
> everything.
> 
> Fixes: aa9aa1ec2df6 ("ARM: davinci: PM: rework init, remove platform device")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> ---
> This patch is just a *guess*. The end of the function looks more like a
> error handling code rather than a normal path.

The patch is correct.

> ---
>  arch/arm/mach-davinci/pm.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/arm/mach-davinci/pm.c b/arch/arm/mach-davinci/pm.c
> index d282b0783ecf..163d865abbf9 100644
> --- a/arch/arm/mach-davinci/pm.c
> +++ b/arch/arm/mach-davinci/pm.c
> @@ -161,6 +161,7 @@ int __init davinci_pm_init(void)
>  						davinci_cpu_suspend_sz);
>  
>  	suspend_set_ops(&davinci_pm_ops);
> +	return 0;

We normally have an empty line before return. I added that while 
applying. Also, I modified the patch description to be more "sure" in 
its language. Here is the final description:

    ARM: davinci: PM: Do not free useful resources in normal path in 'davinci_pm_init'
    
    It is wrong to iounmap resources in the normal path of davinci_pm_init()
    
    The 3 ioremap'ed fields of 'pm_config' can be accessed later on in other
    functions, so we should return 'success' instead of unrolling everything.
    
    Fixes: aa9aa1ec2df6 ("ARM: davinci: PM: rework init, remove platform device")
    Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
    [nsekhar@ti.com: commit message and minor style fixes]
    Signed-off-by: Sekhar Nori <nsekhar@ti.com>

I will let this go through some testing before sending for inclusion in 
v4.12 kernel. Thanks for catching these.

For future, when sending more than one patch, please add a cover 
letter. And have the patches appear as replies to the cover letter. 
This happens if you send the patches together with cover letter using 
git-send-email and have chainreplyto set to false in your .gitconfig

Thanks,
Sekhar