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;
MIME-Version: 1.0
References: <000000000000295a6a05b65efe35@google.com> <00000000000029363505c748757d@google.com>
 <CAJfpegsyFb3bC=dqUbqhs9055TW95EJO2st7iS-4sPME7Y-cmA@mail.gmail.com>
 <CACT4Y+Zh9Bw8DTZyvoOB_hjXwRQuRN+VHmJ-HfMqOaOu7GyVXA@mail.gmail.com> <CAJfpegu2fR3cP+eR24a1+WBBR=hvz8p2cTfK0x4sNz5c-MMH3Q@mail.gmail.com>
In-Reply-To: <CAJfpegu2fR3cP+eR24a1+WBBR=hvz8p2cTfK0x4sNz5c-MMH3Q@mail.gmail.com>
From:   Dmitry Vyukov <dvyukov@google.com>
Date:   Fri, 6 Aug 2021 13:09:47 +0200
Message-ID: <CACT4Y+bh6TGBJAwR5UjoPkCLgiSUypUBWVVMcp9bVEWJbFrASA@mail.gmail.com>
Subject: Re: [syzbot] INFO: task hung in fuse_simple_request
To:     Miklos Szeredi <miklos@szeredi.hu>
Cc:     syzbot <syzkaller@googlegroups.com>,
        syzbot <syzbot+46fe899420456e014d6b@syzkaller.appspotmail.com>,
        areeba.ahmed@futuregulfpak.com, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        syzkaller-bugs <syzkaller-bugs@googlegroups.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, 6 Aug 2021 at 11:34, Miklos Szeredi <miklos@szeredi.hu> wrote:
> > > Hi,
> > >
> > > I'm not sure what to make of this report.
> > >
> > > Does syzbot try to kill all tasks before generating such a hung task report?
> > >
> > > Does it use the fuse specific /sys/fs/fuse/connections/*/abort to
> > > terminate stuck filesystems?
> >
> > Hi Miklos,
> >
> > Grepping the C reproducer for "/sys/fs/fuse/connections", it seems
> > that it tries to abort all connections.
>
> Hi  Dmitry,
>
> So this could indicate a bug in the abort code of fuse, or it could
> indicate that the test case DoS-es the system so that the abort code
> simply doesn't get the resources to execute.
>
> Can the two be differentiated somehow?

I think we need to take a closer look at what happens here. I've filed
https://github.com/google/syzkaller/issues/498#issuecomment-894185550
for this.
The overall idea is that DoS should not be possible by construction
with various sandboxing so that the system does not report what is not
a bug at all. But it's not always easy to achieve.