Re: [PATCH 1/2] platform/x86: silead_dmi - do not treat all devices as i2c_clients

[Dmitry Torokhov]

On Tue, Apr 04, 2017 at 07:33:30PM +0300, Andy Shevchenko wrote:
> On Tue, Apr 4, 2017 at 7:31 PM, Darren Hart <dvhart@xxxxxxxxxxxxx> wrote:
> > On Tue, Apr 04, 2017 at 07:08:28PM +0300, Andy Shevchenko wrote:
> >> On Tue, Apr 4, 2017 at 7:05 PM, Darren Hart <dvhart@xxxxxxxxxxxxx> wrote:
> >> > On Tue, Apr 04, 2017 at 04:45:05PM +0300, Andy Shevchenko wrote:
> >> >> On Tue, Apr 4, 2017 at 6:25 AM, Dmitry Torokhov
> >> >> <dmitry.torokhov@xxxxxxxxx> wrote:
> >> >> > I2C bus has both i2c clients and adapter devices, so we must be careful in
> >> >> > notifier code and verify that we are actually dealing with an i2c client
> >> >> > before using it as such.
> >> >>
> >> >> > -static void silead_ts_dmi_add_props(struct device *dev)
> >> >> > +static void silead_ts_dmi_add_props(struct i2c_client *client)
> >> >> >  {
> >> >>
> >> >> > -       struct i2c_client *client = to_i2c_client(dev);
> >> >>
> >> >> I would replace this by
> >> >> struct device *dev = &client->dev;
> >> >>
> >> >> Otherwise looks good for me.
> >> >
> >> > Andy, this series looks like a candidate for 4.11-fixes. We're already at rc5
> >> > though, so if we are going to do that, I'd like to see a stronger statement in
> >> > the commit log about how this issue manifests currently - if it does.
> >>
> >> It makes less changes for any (potentially) backported code.
> >> I'm not insisting and even can do myself.

OK, I'll leave it to you if you decide you'd rather have a temporary for
'dev'.

> >
> > Sorry, I was referring to the series itself, not your feedback above. You
> > assigned this to yourself in patchwork, so I was just noting that this patch
> > series may be a candidate for fixes to 4.11, rather than testing/for-next for
> > 4.12. Your call.
> 
> Ah, thanks Darren for clarification.

I think this can easily wait for 4.12 if you are concerned about
regressions. i2c_adapter is large enough so we won't reference
unallocated memory and we only read from client->name so we should not
perturb anything if we actually are dealing with i2c_adapter, and the
data we'll read is unlikely to match to ACPI name we interested in.

The 2nd patch is mere optimization.

Thanks.

-- 
Dmitry