Re: [PATCH] sched: Memory leak in two error corner cases

[Peter Zijlstra <peterz@infradead.org>]

On Wed, 2009-12-09 at 19:45 +0200, Phil Carmody wrote:
> From: Phil Carmody <ext-phil.2.carmody@nokia.com>
> 
> If the second in each of these pairs of allocations fails, then
> the first one will not be freed in the error route out.
> 
> Found by a static code analysis tool.

nice find.

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>

> Signed-off-by: Phil Carmody <ext-phil.2.carmody@nokia.com>
> ---
>  kernel/sched.c |    8 ++++++--
>  1 files changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/sched.c b/kernel/sched.c
> index e7f2cfa..29ebc4a 100644
> --- a/kernel/sched.c
> +++ b/kernel/sched.c
> @@ -9841,8 +9841,10 @@ int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
>  
>  		se = kzalloc_node(sizeof(struct sched_entity),
>  				  GFP_KERNEL, cpu_to_node(i));
> -		if (!se)
> +		if (!se) {
> +			kfree(cfs_rq);
>  			goto err;
> +		}
>  
>  		init_tg_cfs_entry(tg, cfs_rq, se, i, 0, parent->se[i]);
>  	}
> @@ -9929,8 +9931,10 @@ int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
>  
>  		rt_se = kzalloc_node(sizeof(struct sched_rt_entity),
>  				     GFP_KERNEL, cpu_to_node(i));
> -		if (!rt_se)
> +		if (!rt_se) {
> +			kfree(rt_rq);
>  			goto err;
> +		}
>  
>  		init_tg_rt_entry(tg, rt_rq, rt_se, i, 0, parent->rt_se[i]);
>  	}