Received-SPF: pass (google.com: domain of linux-kernel+bounces-101636-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Precedence: bulk
MIME-Version: 1.0
References: <03d7a29c7e1a8c5741680ea9bc83b4fb40358a25.camel@elektrobit.com> <CAOQ4uxg+RveBHjgui_HjCasYGor3JNeuv-UroR=5j4n6TgRd7w@mail.gmail.com>
In-Reply-To: <CAOQ4uxg+RveBHjgui_HjCasYGor3JNeuv-UroR=5j4n6TgRd7w@mail.gmail.com>
From: Miklos Szeredi <miklos@szeredi.hu>
Date: Wed, 13 Mar 2024 15:47:17 +0100
Message-ID: <CAJfpeguBfKB4MGEUkJ=+ZFT9uYJ1DXtCTW9oLrEE0af7XTi=VA@mail.gmail.com>
Subject: Re: possible deadlock in ovl_llseek 27c1936af506
To: Amir Goldstein <amir73il@gmail.com>
Cc: =?UTF-8?Q?Wei=C3=9F=2C_Simone?= <Simone.Weiss@elektrobit.com>, 
	"linux-unionfs@vger.kernel.org" <linux-unionfs@vger.kernel.org>, 
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"

On Wed, 13 Mar 2024 at 14:14, Amir Goldstein <amir73il@gmail.com> wrote:

> The reason for this report is calling llseek() on lower ovl from
> ovl_copy_up_data() when ovl_copy_up_data() is called with upper
> inode lock and the lower ovl uses the same upper fs.
>
> It looks to me like the possible deadlock should have been solved by commit
> c63e56a4a652 ovl: do not open/llseek lower file with upper sb_writers held
> that moved ovl_copy_up_data() out of the inode_lock() scope.

That commit is in v6.7, so something different must be happening on v6.8-rc1.

Simone, please send a new report for v6.8-rc1 if a lockdep splat can
be reproduced on that kernel.

Thanks,
Miklos