On Tue, Apr 02, 2019 at 03:47:47PM +0300, Heikki Krogerus wrote:
> On Mon, Apr 01, 2019 at 02:45:11PM +0200, Hans de Goede wrote:
> > HI,
> >
> > On 01-04-19 14:40, Heikki Krogerus wrote:
> > > On Mon, Apr 01, 2019 at 12:34:29PM +0200, Greg KH wrote:
> > > > On Mon, Apr 01, 2019 at 01:15:53PM +0300, Heikki Krogerus wrote:
> > > > > Registering real device entries (struct device) for the mode
> > > > > muxes as well as for the orientation switches.
> > > > >
> > > > > The Type-C mux code was deliberately attempting to avoid
> > > > > creation of separate device entries for the orientation
> > > > > switch and the mode switch (alternate modes) because they
> > > > > are not physical devices. They are functions of a single
> > > > > physical multiplexer/demultiplexer switch device.
> > > > >
> > > > > Unfortunately because of the dependency we still have on the
> > > > > underlying mux device driver, we had to put in hacks like
> > > > > the one in the commit 3e3b81965cbf ("usb: typec: mux: Take
> > > > > care of driver module reference counting") to make sure the
> > > > > driver does not disappear from underneath us. Even with
> > > > > those hacks we were still left with a potential NUll pointer
> > > > > dereference scenario, so just creating the device entries,
> > > > > and letting the core take care of the dependencies. No more
> > > > > hacks needed.
> > > > >
> > > > > Fixes: 3e3b81965cbf ("usb: typec: mux: Take care of driver module reference counting")
> > > > > Cc: v4.19.x <stable@xxxxxxxxxxxxxxx> # v4.19.x+
> > > > > Signed-off-by: Heikki Krogerus <heikki.krogerus@xxxxxxxxxxxxxxx>
> > > >
> > > > This looks good to me, nice work!
> > > >
> > > > But, it would be nice if someone who has this hardware can test it to
> > > > verify it does actually work :)
> > >
> > > This alone does not work on Intel Cherrytrail platforms. I need to make
> > > the Intel Cherrytrail MFD driver (intel_cht_int33fe.c) to use the new
> > > device names that we now have for the muxes. Sorry for the mistake.
> > >
> > > I'll resend this and include the needed modifications to
> > > intel_cht_int33fe.c. Hans should be able to test this once I do that. I
> > > hope he has time.
> >
> > Yes I need to get back to the CherryTrail Type-C stuff we discussed a while
> > back anyways. On which tree/branch should I test v2 of this patch(series) ?
>
> Greg's usb-next.
I just realized that the potential NULL pointer dereference that this
patch is meant to fix can't happen until my "software node reference"
modifications [1] have been applied. I don't think this needs to be
considered a fix.
I think the correct way to handle this patch is to make just it part
of that series. I'll resend the series, and include this patch in it.
[1] https://lkml.org/lkml/2019/3/15/457
thanks,
--
heikki
