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: <5d5ccbb1-784c-52b3-3748-2cf7b5cf01ef@kernel.dk>
Date:   Tue, 13 Jun 2023 19:14:25 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
Subject: Re: [PATCH] io_uring/io-wq: don't clear PF_IO_WORKER on exit
Content-Language: en-US
To:     Zorro Lang <zlang@redhat.com>
Cc:     io-uring <io-uring@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Dave Chinner <david@fromorbit.com>
References: <2392dcb4-71f4-1109-614b-4e2083c0941e@kernel.dk>
 <20230614005449.awc2ncxl5lb2eg6m@zlang-mailbox>
From:   Jens Axboe <axboe@kernel.dk>
In-Reply-To: <20230614005449.awc2ncxl5lb2eg6m@zlang-mailbox>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 6/13/23 6:54?PM, Zorro Lang wrote:
> On Mon, Jun 12, 2023 at 12:11:57PM -0600, Jens Axboe wrote:
>> A recent commit gated the core dumping task exit logic on current->flags
>> remaining consistent in terms of PF_{IO,USER}_WORKER at task exit time.
>> This exposed a problem with the io-wq handling of that, which explicitly
>> clears PF_IO_WORKER before calling do_exit().
>>
>> The reasons for this manual clear of PF_IO_WORKER is historical, where
>> io-wq used to potentially trigger a sleep on exit. As the io-wq thread
>> is exiting, it should not participate any further accounting. But these
>> days we don't need to rely on current->flags anymore, so we can safely
>> remove the PF_IO_WORKER clearing.
>>
>> Reported-by: Zorro Lang <zlang@redhat.com>
>> Reported-by: Dave Chinner <david@fromorbit.com>
>> Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
>> Link: https://lore.kernel.org/all/ZIZSPyzReZkGBEFy@dread.disaster.area/
>> Fixes: f9010dbdce91 ("fork, vhost: Use CLONE_THREAD to fix freezer/ps regression")
>> Signed-off-by: Jens Axboe <axboe@kernel.dk>
>>
>> ---
> 
> Hi,
> 
> This patch fix the issue I reported. The bug can be reproduced on v6.4-rc6,
> then test passed on v6.4-rc6 with this patch.
> 
> But I found another KASAN bug [1] on aarch64 machine, by running generic/388.
> I hit that 3 times. And hit a panic [2] (once after that kasan bug) on a x86_64
> with pmem device (mount with dax=never), by running geneirc/388 too.

Can you try with this? I suspect the preempt dance isn't really
necessary, but I can't quite convince myself that it isn't. In any case,
I think this should fix it and this was exactly what I was worried about
but apparently not able to easily trigger or prove...


diff --git a/io_uring/io-wq.c b/io_uring/io-wq.c
index fe38eb0cbc82..878ec3feeba9 100644
--- a/io_uring/io-wq.c
+++ b/io_uring/io-wq.c
@@ -220,7 +220,9 @@ static void io_worker_exit(struct io_worker *worker)
 	list_del_rcu(&worker->all_list);
 	raw_spin_unlock(&wq->lock);
 	io_wq_dec_running(worker);
-	worker->flags = 0;
+	preempt_disable();
+	current->worker_private = NULL;
+	preempt_enable();
 
 	kfree_rcu(worker, rcu);
 	io_worker_ref_put(wq);

-- 
Jens Axboe