Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Waiman Long <llong@redhat.com>
Subject: Re: [PATCH] cgroup/debug: Fix lockdep splat with "%pK" format
 specifier
To:     Tejun Heo <tj@kernel.org>
Cc:     Zefan Li <lizefan.x@bytedance.com>,
        Johannes Weiner <hannes@cmpxchg.org>, cgroups@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20210927025807.26918-1-longman@redhat.com>
 <YVH0uKGGmRjrIlFy@slm.duckdns.org>
Message-ID: <4f4f5a8c-7488-5ea6-0f1b-6ac95c259b4a@redhat.com>
Date:   Mon, 27 Sep 2021 14:17:47 -0400
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <YVH0uKGGmRjrIlFy@slm.duckdns.org>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Content-Language: en-US
Precedence: bulk

On 9/27/21 12:43 PM, Tejun Heo wrote:
> Hello,
>
> On Sun, Sep 26, 2021 at 10:58:07PM -0400, Waiman Long wrote:
>> The lockdep splat is caused by the fact that the debug controller use the
>> "%pK" format specifier to print out address of cset's with css_set_lock
>> held. Under some circumstances, the use of "%pK" format specifier may
>> acquire the selinux_ss.policy_rwlock.
>>
>> To avoid this possible deadlock scenario, we have to abandon the use of
>> the "%pK" format specifier and just use "%p" to always hash the cset
>> addresses. The actual cset addresses aren't that important as long as
>> they are unique for matching purpose.
> Isn't the right thing to do here making the selinux rwlock an irqsafe one?
> It's a bit crazy to have printf specifier to have restrictive locking
> requirements.

I just realize that upstream had get rid of the policy_rwlock since 
v5.10 and use RCU instead. So this patch isn't applicable then.

My original thought was that policy_rwlock were referenced over 60 times 
in security/selinux/ss/services.c whereas we only use a few %pK format 
specifier in the debug controller. Anyway, it is moot now.

Sorry for the noise.

Cheers,
Longman