Received-SPF: pass (google.com: domain of linux-kernel+bounces-153038-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
From: Jeongjun Park <aha310510@gmail.com>
To: jirislaby@kernel.org
Cc: davem@davemloft.net,
	edumazet@google.com,
	gregkh@linuxfoundation.org,
	kuba@kernel.org,
	linux-kernel@vger.kernel.org,
	linux-ppp@vger.kernel.org,
	netdev@vger.kernel.org,
	pabeni@redhat.com,
	syzbot+393d0ef63475d9bb1f16@syzkaller.appspotmail.com,
	syzkaller-bugs@googlegroups.com
Subject: Re: [PATCH net,v3] net: ppp: Fix deadlock caused by unsafe-irq lock in ap_get()
Date: Mon, 22 Apr 2024 17:58:01 +0900
Message-Id: <20240422085801.140509-1-aha310510@gmail.com>
In-Reply-To: <bb5c1fd1-c3e7-4aaf-b28d-a45479ff0815@kernel.org>
References: <bb5c1fd1-c3e7-4aaf-b28d-a45479ff0815@kernel.org>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

Jiri Slaby wrote:
> OK, but read_unlock_irq() is not the right way to do it. It would enable
> irqs unconditionally, so serial8250_tx_chars() would not thank you in
> the above call chain.
> 
> Plus the issues Jakub had with the patch.
> 
> And what about ppp_synctty?

Hmm... If the problem is the case of calling the function in 
serial8250_tx_chars(), I think it can be solved by continuing to use 
read_lock() in ppp_asynctty_wakeup() and getting the ap by not calling
ap_get().

If I do this, will I be able to apply the existing patch to ap_get()?

Thanks.