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: <de6b0f81-67e1-b167-a618-193c2303f149@kernel.dk>
Date:   Wed, 18 May 2022 16:31:53 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.1
Subject: Re: [PATCH] blk-cgroup: Remove unnecessary rcu_read_lock/unlock()
Content-Language: en-US
From:   Jens Axboe <axboe@kernel.dk>
To:     Marek Szyprowski <m.szyprowski@samsung.com>, bh1scw@gmail.com,
        tj@kernel.org
Cc:     cgroups@vger.kernel.org, linux-block@vger.kernel.org,
        linux-kernel@vger.kernel.org, songmuchun@bytedance.com
References: <20220516173930.159535-1-bh1scw@gmail.com>
 <CGME20220518192850eucas1p1458c00d4917c5ed39f2c37c9eb30cd46@eucas1p1.samsung.com>
 <46253c48-81cb-0787-20ad-9133afdd9e21@samsung.com>
 <1dad86bb-ae31-5bf8-5810-9e81c68be8ff@kernel.dk>
In-Reply-To: <1dad86bb-ae31-5bf8-5810-9e81c68be8ff@kernel.dk>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 5/18/22 4:29 PM, Jens Axboe wrote:
> On 5/18/22 1:28 PM, Marek Szyprowski wrote:
>> On 16.05.2022 19:39, bh1scw@gmail.com wrote:
>>> From: Fanjun Kong <bh1scw@gmail.com>
>>>
>>> spin_lock_irq/spin_unlock_irq contains preempt_disable/enable().
>>> Which can serve as RCU read-side critical region, so remove
>>> rcu_read_lock/unlock().
>>>
>>> Signed-off-by: Fanjun Kong <bh1scw@gmail.com>
>>
>> This patch landed in today's linux next-20220518 as commit 77c570a1ea85 
>> ("blk-cgroup: Remove unnecessary rcu_read_lock/unlock()").
>>
>> Unfortunately it triggers the following warning on ARM64 based Raspberry 
>> Pi 4B board:>
>> ------------[ cut here ]------------
>> WARNING: CPU: 0 PID: 1 at block/blk-cgroup.c:301 blkg_create+0x398/0x4e0
> 
> Should this use rcu_read_lock_any_held() rather than
> rcu_read_lock_held()?

I think the better alternative is just to delete the WARN_ON(), we have
a:

lockdep_assert_held(&q->queue_lock);

right after it. Since the queue_lock is IRQ disabling, having two checks
serves no purpose. I'll kill the line.

-- 
Jens Axboe