Re: [PATCH] blk-mq: avoid stall during boot due to synchronize_rcu_expedited

[Jens Axboe <axboe@kernel.dk>]

On 1/6/26 9:59 AM, Mikulas Patocka wrote:
> @@ -4553,8 +4566,14 @@ static void __blk_mq_realloc_hw_ctxs(str
>  		 * Make sure reading the old queue_hw_ctx from other
>  		 * context concurrently won't trigger uaf.
>  		 */
> -		synchronize_rcu_expedited();
> -		kfree(hctxs);
> +		r = kmalloc(sizeof(struct rcu_free_hctxs), GFP_KERNEL);
> +		if (!r) {
> +			synchronize_rcu_expedited();
> +			kfree(hctxs);
> +		} else {
> +			r->hctxs = hctxs;
> +			call_rcu(&r->head, rcu_free_hctxs);
> +		}
>  		hctxs = new_hctxs;
>  	}

This is worse in every conceivable way, imho. The proper way to do this
would be to embed the rcu_head in whatever is allocated for the hctxs at
alloc time, if youre doing an alloc here you may as well just use
kfree_rcu_mightsleep() in the first place. There's nothing gained from
open coding that.

Since kfree_rcu_mightsleep() will only run into trouble under strained
conditions anyway, I think the original patch is fine for this.

-- 
Jens Axboe