Re: [PATCH] io: adc: stm32-adc: fix potential NULL pointer dereference in stm32_adc_probe()

[Zhang Shurong <zhang_shurong@foxmail.com>]

在 2023年8月29日星期二 CST 上午12:16:05，Jonathan Cameron 写道：
> On Mon, 28 Aug 2023 23:02:07 +0800
> 
> Zhang Shurong <zhang_shurong@foxmail.com> wrote:
> > 在 2023年7月17日星期一 CST 上午12:08:21，Jonathan Cameron 写道：
> > 
> > > On Sat, 15 Jul 2023 23:55:50 +0800
> > > 
> > > Zhang Shurong <zhang_shurong@foxmail.com> wrote:
> > > > of_match_device() may fail and returns a NULL pointer.
> > > > 
> > > > Fix this by checking the return value of of_match_device().
> > > > 
> > > > Fixes: 64ad7f6438f3 ("iio: adc: stm32: introduce compatible data cfg")
> > > > Signed-off-by: Zhang Shurong <zhang_shurong@foxmail.com>
> > > 
> > > Hi Zhang,
> > > 
> > > I'm not sure we can actually make this bug happen. Do you have
> > > a way of triggering it?  The driver is only probed on devices where
> > > that match will work.
> > > 
> > > Also, assuming the match table is the same one associated with this
> > > probe
> > > function, then us priv->cfg = of_device_get_match_data() and check the
> > > output of that which is what we really care about.
> > > 
> > > Jonathan
> > > 
> > > > ---
> > > > 
> > > >  drivers/iio/adc/stm32-adc-core.c | 9 +++++++--
> > > >  1 file changed, 7 insertions(+), 2 deletions(-)
> > > > 
> > > > diff --git a/drivers/iio/adc/stm32-adc-core.c
> > > > b/drivers/iio/adc/stm32-adc-core.c index 48f02dcc81c1..70011fdbf5f6
> > > > 100644
> > > > --- a/drivers/iio/adc/stm32-adc-core.c
> > > > +++ b/drivers/iio/adc/stm32-adc-core.c
> > > > @@ -706,6 +706,8 @@ static int stm32_adc_probe(struct platform_device
> > > > *pdev)>
> > > > 
> > > >  	struct stm32_adc_priv *priv;
> > > >  	struct device *dev = &pdev->dev;
> > > >  	struct device_node *np = pdev->dev.of_node;
> > > > 
> > > > +	const struct of_device_id *of_id;
> > > > +
> > > > 
> > > >  	struct resource *res;
> > > >  	u32 max_rate;
> > > >  	int ret;
> > > > 
> > > > @@ -718,8 +720,11 @@ static int stm32_adc_probe(struct platform_device
> > > > *pdev)>
> > > > 
> > > >  		return -ENOMEM;
> > > >  	
> > > >  	platform_set_drvdata(pdev, &priv->common);
> > > > 
> > > > -	priv->cfg = (const struct stm32_adc_priv_cfg *)
> > > > -		of_match_device(dev->driver->of_match_table, dev)->data;
> > > > +	of_id = of_match_device(dev->driver->of_match_table, dev);
> > > > +	if (!of_id)
> > > > +		return -ENODEV;
> > > > +
> > > > +	priv->cfg = (const struct stm32_adc_priv_cfg *)of_id->data;
> > > > 
> > > >  	priv->nb_adc_max = priv->cfg->num_adcs;
> > > >  	spin_lock_init(&priv->common.lock);
> > 
> > Hello Jonathan,
> > 
> > I think we can make it happen by designing the platform device in a way
> > that its name aligns with that of the driver. In such a scenario, when
> > the driver is probed, the of_match_device function will return null. You
> > can verify this functionality by reviewing the following function:
> > 
> > static int platform_match(struct device *dev, struct device_driver *drv)
> 
> I don't think we care about that case.  If there is a real example of
> why someone would do this then that would be a different matter.
> 
> Jonathan
> 
> > Best regards,
> > Shurong
I just think it might be more appropriate to return the correct error code in 
this situation. I agree with your assessment that it is an abnormal case. 
Therefore, it is perfectly fine if you decide not to select this patch.

Thanks for your kind reply.

Shurong