[PATCH] genalloc: fix device node resource counter

[PATCH] genalloc: fix device node resource counter

[Vladimir Zapolskiy]

The change balances a usage counter of np_pool device_node, which is
incremented on preceding of_parse_phandle() call.

Signed-off-by: Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Olof Johansson <olof@lixom.net>
Cc: Catalin Marinas <catalin.marinas@arm.com>
---
 lib/genalloc.c |    1 +
 1 file changed, 1 insertion(+)

diff --git a/lib/genalloc.c b/lib/genalloc.c
index c7a91cf..cce4dd6 100644
--- a/lib/genalloc.c
+++ b/lib/genalloc.c
@@ -637,6 +637,7 @@ struct gen_pool *of_get_named_gen_pool(struct device_node *np,
 	if (!np_pool)
 		return NULL;
 	pdev = of_find_device_by_node(np_pool);
+	of_node_put(np_pool);
 	if (!pdev)
 		return NULL;
 	return dev_get_gen_pool(&pdev->dev);
-- 
1.7.10.4

Re: [PATCH] genalloc: fix device node resource counter

[Andrew Morton]

On Wed, 24 Sep 2014 22:51:00 +0300 Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com> wrote:

> The change balances a usage counter of np_pool device_node, which is
> incremented on preceding of_parse_phandle() call.
> 
> ...
>
> --- a/lib/genalloc.c
> +++ b/lib/genalloc.c
> @@ -637,6 +637,7 @@ struct gen_pool *of_get_named_gen_pool(struct device_node *np,
>  	if (!np_pool)
>  		return NULL;
>  	pdev = of_find_device_by_node(np_pool);
> +	of_node_put(np_pool);
>  	if (!pdev)
>  		return NULL;
>  	return dev_get_gen_pool(&pdev->dev);

Looks good, thanks.

However...  when fixing a bug, please always describe the end-user
visible effects of that bug.  Amongst other things this is to help me
and others decide which kernel version(s) need the patch.

I'm assuming that the effect will be, at worst, a memory leak at the
rate of one device_node per hot-unplug event and is hence a very minor
thing?

Re: [PATCH] genalloc: fix device node resource counter

[Vladimir Zapolskiy]

Hi Andrew,

On 24.09.2014 23:58, Andrew Morton wrote:
> On Wed, 24 Sep 2014 22:51:00 +0300 Vladimir Zapolskiy <vladimir_zapolskiy@mentor.com> wrote:
>
>> The change balances a usage counter of np_pool device_node, which is
>> incremented on preceding of_parse_phandle() call.
>>
>> ...
>>
>> --- a/lib/genalloc.c
>> +++ b/lib/genalloc.c
>> @@ -637,6 +637,7 @@ struct gen_pool *of_get_named_gen_pool(struct device_node *np,
>>   	if (!np_pool)
>>   		return NULL;
>>   	pdev = of_find_device_by_node(np_pool);
>> +	of_node_put(np_pool);
>>   	if (!pdev)
>>   		return NULL;
>>   	return dev_get_gen_pool(&pdev->dev);
>
> Looks good, thanks.
>
> However...  when fixing a bug, please always describe the end-user
> visible effects of that bug.  Amongst other things this is to help me
> and others decide which kernel version(s) need the patch.
>
> I'm assuming that the effect will be, at worst, a memory leak at the
> rate of one device_node per hot-unplug event and is hence a very minor
> thing?
>

yes, you are correct, moreover there are not so many users of 
of_get_named_gen_pool() in vanilla (I counted 3), however since the 
change is in ./lib folder, it may be considered as relatively important 
one. I think it is sufficient to have the change in linux-next only.

With best wishes,
Vladimir