Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <330468da-a778-6cfc-ce46-a62566d71c8d@kernel.dk>
Date:   Fri, 12 Aug 2022 06:40:32 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:91.0) Gecko/20100101
 Thunderbird/91.10.0
Subject: Re: [PATCH] blk-mq: move bio merge attempt to blk_mq_submit_bio()
Content-Language: en-US
To:     Chengming Zhou <zhouchengming@bytedance.com>
Cc:     linux-block@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20220812083944.79616-1-zhouchengming@bytedance.com>
From:   Jens Axboe <axboe@kernel.dk>
In-Reply-To: <20220812083944.79616-1-zhouchengming@bytedance.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 8/12/22 2:39 AM, Chengming Zhou wrote:
> We will try to get request from cache before alloc request, in
> both cases will first attempt bio merge.
> 
> This patch move this common part to blk_mq_submit_bio(), which
> simplify the code and avoid passing in the pointer of bio.

I don't think this will work. If we get a cached request, then we
know we have a reference to the queue. If you swap these around,
then that's no longer the case and we can have a use-after-free if
the queue goes away.

-- 
Jens Axboe