[Devel] Re: [PATCH] Fix ckpt_obj_lookup_add() leak detection logic

[Oren Laadan]

Got it, thanks.

Dan Smith wrote:
> Since first is a pointer, it's never NULL and thus leak detection will
> always be performed (of CHECKPOINT_SUBTREE is set).  This fatally breaks
> the checkpoint process when that flag is enabled because any object (with
> a ref_users operation) that is added to the hash twice will be reported as
> a reverse leak.
> 
> Signed-off-by: Dan Smith <danms at us.ibm.com>
> ---
>  checkpoint/objhash.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/checkpoint/objhash.c b/checkpoint/objhash.c
> index 8dfb0b0..b85aa77 100644
> --- a/checkpoint/objhash.c
> +++ b/checkpoint/objhash.c
> @@ -679,7 +679,7 @@ int ckpt_obj_lookup_add(struct ckpt_ctx *ctx, void *ptr,
>  	ckpt_debug("%s objref %d first %d\n",
>  		   obj->ops->obj_name, obj->objref, *first);
>  
> -	if (first && obj_reverse_leak(ctx, obj))
> +	if (*first && obj_reverse_leak(ctx, obj))
>  		return -EBUSY;
>  
>  	obj->flags |= CKPT_OBJ_VISITED;
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers