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: <bf5c8a1e-e922-fb78-32b8-a6288e434a3a@I-love.SAKURA.ne.jp>
Date:   Mon, 16 May 2022 10:56:58 +0900
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.9.0
Subject: Re: [PATCH v3 (repost)] workqueue: Warn flushing of kernel-global
 workqueues
Content-Language: en-US
To:     Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc:     Linus Torvalds <torvalds@linux-foundation.org>,
        Lai Jiangshan <jiangshanlai@gmail.com>,
        LKML <linux-kernel@vger.kernel.org>, Tejun Heo <tj@kernel.org>
References: <2efd5461-fccd-f1d9-7138-0a6767cbf5fe@I-love.SAKURA.ne.jp>
 <CAHk-=wgiCVcF4vJCOcg1hDw9PMT6zNSgt-pPuZL8ihWEcVv3eg@mail.gmail.com>
 <e4203c4f-9c3d-6e33-06e8-052676cc5af1@I-love.SAKURA.ne.jp>
 <YjivtdkpY+reW0Gt@slm.duckdns.org> <YnzjrnnP622mad1y@google.com>
 <33772dcb-c5dd-c5e7-30c8-d2397d21b26c@I-love.SAKURA.ne.jp>
 <Yn0H/68tagxaj/ke@google.com>
From:   Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
In-Reply-To: <Yn0H/68tagxaj/ke@google.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2022/05/12 22:13, Dmitry Torokhov wrote:
> Also as far as I can see the patch was rejected.

Not rejected. I sent
https://lkml.kernel.org/r/7b2fecdb-59ae-2c54-5a5b-774ef7054d1b@I-love.SAKURA.ne.jp
for 5.19.

>> We currently don't have a flag to tell whether the caller is inside module unload
>> path. And even inside module unload path, flushing the system-wide workqueue is
>> problematic under e.g. GFP_NOFS/GFP_NOIO context.
> 
> Sorry, I do not follow here. Are there module unloading code that runs
> as GFP_NOFS/GFP_NOIO?

It is GFP_KERNEL context when module's __exit function is called. But whether
flush_workqueue() is called from restricted context depends on what locks does
the module's __exit function hold.

If request_module() is called from some work function using one of system-wide WQs,
and flush_workqueue() is called on that WQ from module's __exit function, the kernel
might deadlock on module_mutex lock. Making sure that flush_workqueue() is not called
on system-wide WQs is the safer choice.