Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Message-ID: <1565101558.8136.30.camel@suse.com>
Subject: Re: KASAN: use-after-free Read in device_release_driver_internal
From:   Oliver Neukum <oneukum@suse.com>
To:     Alan Stern <stern@rowland.harvard.edu>
Cc:     Andrey Konovalov <andreyknvl@google.com>,
        syzkaller-bugs <syzkaller-bugs@googlegroups.com>,
        syzbot <syzbot+1b2449b7b5dc240d107a@syzkaller.appspotmail.com>,
        LKML <linux-kernel@vger.kernel.org>,
        USB list <linux-usb@vger.kernel.org>
Date:   Tue, 06 Aug 2019 16:25:58 +0200
In-Reply-To: <Pine.LNX.4.44L0.1908061009450.1571-100000@iolanthe.rowland.org>
References: <Pine.LNX.4.44L0.1908061009450.1571-100000@iolanthe.rowland.org>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Am Dienstag, den 06.08.2019, 10:19 -0400 schrieb Alan Stern:
> On Tue, 6 Aug 2019, Oliver Neukum wrote:
> 
> > Am Donnerstag, den 01.08.2019, 14:47 -0400 schrieb Alan Stern:
> > > 
> > > I think this must be caused by an unbalanced refcount.  That is,
> > > something must drop one more reference to the device than it takes.
> > > That would explain why the invalid access occurs inside a single
> > > bus_remove_device() call, between the klist_del() and
> > > device_release_driver().
> > > 
> > > The kernel log indicates that the device was probed by rndis_wlan,
> > > rndis_host, and cdc_acm, all of which got errors because of the
> > > device's bogus descriptors.  Probably one of them is messing up the
> > > refcount.
> > 
> > Hi,
> > 
> > you made me look at cdc-acm. I suspect
> > 
> > cae2bc768d176bfbdad7035bbcc3cdc973eb7984 ("usb: cdc-acm: Decrement tty port's refcount if probe() fail")
> > 
> > is buggy decrementing the refcount on the interface in destroy()
> > even before the refcount is increased.
> > 
> > Unfortunately I cannot tell from the bug report how many and which
> > interfaces the emulated test device has. Hence it is unclear to me,
> > when exactly probe() would fail cdc-acm.
> 
> Only one interface (numbered 234!).

Yes. cdc-acm went into the look_for_collapsed_interface code path.
But I cannot tell whether it proceeded to made_compressed_probe

(Yes, I know the code makes extensive use of "goto")

> > If you agree. I am attaching a putative fix.
> 
> Your patch adds a line saying:
> 
> > +	usb_get_intf(acm->control); /* undone in destroy() */
> 
> but I don't see any destroy() function in that source file.  Did you 
> mean acm_port_destruct()?

Yes, sorry

> In any case, I don't know if this missing "get" would cause the 
> problem, but it might well.

Then let's wait for the result.

	Regards
		Oliver