Re: [PATCH v2] blk-cgroup: remove entries in blkg_tree before queue release

[Alexandru Moise]

On Wed, Apr 11, 2018 at 01:55:25PM -0600, Bart Van Assche wrote:
> On 04/11/18 13:00, Alexandru Moise wrote:
> > But the root cause of it is in blkcg_init_queue() when blkg_create() returns
> > an ERR ptr, because it tries to insert into a populated index into blkcg->blkg_tree,
> > the entry that we fail to remove at __blk_release_queue().
> 
> Hello Alex,
> 
> Had you considered something like the untested patch below?
> 
> Thanks,
> 
> Bart.
> 
> 
> diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
> index 1c16694ae145..f2ced19e74b8 100644
> --- a/block/blk-cgroup.c
> +++ b/block/blk-cgroup.c
> @@ -1191,14 +1191,17 @@ int blkcg_init_queue(struct request_queue *q)
>  	if (preloaded)
>  		radix_tree_preload_end();
> 
> -	if (IS_ERR(blkg))
> -		return PTR_ERR(blkg);
> +	if (IS_ERR(blkg)) {
> +		ret = PTR_ERR(blkg);
> +		goto destroy_all;
> +	}
> 
>  	q->root_blkg = blkg;
>  	q->root_rl.blkg = blkg;
> 
>  	ret = blk_throtl_init(q);
>  	if (ret) {
> +destroy_all:
>  		spin_lock_irq(q->queue_lock);
>  		blkg_destroy_all(q);
>  		spin_unlock_irq(q->queue_lock);
> 

Hi, I tested it, it doesn't solve the problem.
By the time you get here it's already too late, my patch
prevents this from failing in the first place.

I would have liked this more than my solution though.

../Alex