linux-kernel - Re: [PATCHv11 2/3] usb: USB Type-C connector class

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 21 Nov 2016 15:11:03 +0200
From:   Heikki Krogerus <heikki.krogerus@...ux.intel.com>
To:     Greg KH <gregkh@...uxfoundation.org>
Cc:     Guenter Roeck <linux@...ck-us.net>,
        Badhri Jagan Sridharan <badhri@...gle.com>,
        Oliver Neukum <oneukum@...e.com>, linux-kernel@...r.kernel.org,
        linux-usb@...r.kernel.org
Subject: Re: [PATCHv11 2/3] usb: USB Type-C connector class

Hi Greg,

On Mon, Nov 21, 2016 at 11:35:28AM +0100, Greg KH wrote:
> > +static void typec_partner_release(struct device *dev)
> > +{
> > +	struct typec_port *port = to_typec_port(dev->parent);
> > +
> > +	typec_unregister_altmodes(dev);
> > +	port->partner = NULL;
> > +}
> 
> This doesn't feel right, why are you both exporting
> typec_unregister_altmodes() and also calling it from release callbacks?
> That implies there is two way to clean stuff up, so what is a driver
> writer to use?  Please simplify and force it to be one way or the other.

OK.

> Also typec_unregister_altmodes() doesn't free memory, which release is
> supposed to be doing, which also implies that the reference counting
> logic is all wrong here.  Please fix, like I asked you to previously.

There is nothing wrong with the reference counting, and nothing has
been allocated so there is nothing to free.

Please note that the partner device is meant to just represent the
partner in user space and not to be actually used for anything. And
please also note that there can only be one partner for a port at a
time.

We could allocate an extra structure for the partner when
typec_connect() is called, but we would do that just for the sake of
having something to free in the release hook. It would not be useful
for anything. It would not help us increase/decrease the reference
count of the device, and the port driver would still have to provide
details about the partner capabilities the moment it tells us the
partner was connected.

Thanks,

-- 
heikki