On Mon, Sep 07, 2020 at 03:15:17PM +0200, Tomasz Figa wrote: > On Fri, Sep 4, 2020 at 4:06 PM Andy Shevchenko > <andy.shevchenko@xxxxxxxxx> wrote: > > On Fri, Sep 4, 2020 at 4:48 PM Dongchun Zhu <dongchun.zhu@xxxxxxxxxxxx> wrote: > > > On Wed, 2020-09-02 at 16:44 +0300, Andy Shevchenko wrote: > > > > On Wed, Sep 02, 2020 at 08:01:22PM +0800, Dongchun Zhu wrote: ... > > > > > + struct i2c_client *client = to_i2c_client(dev); > > > > > + struct v4l2_subdev *sd = i2c_get_clientdata(client); > > > > > > > > struct v4l2_subdev *sd = dev_get_drvdata(dev); > > > > > > > > Same for the rest similar cases. > > > > > > We've discussed the issue in DW9768 V2. > > > > > > For V4L2 sub-device drivers, dev_get_drvdata() shouldn't be used > > > directly. > > > > > > More details please check the Google Issue: > > > https://partnerissuetracker.corp.google.com/issues/147957975 > > > > This is not a public link. Can you remind me what was the issue? > > > > v4l2-subdev framework uses dev drvdata for its own purposes. However, > that problem was about the driver setting its own drvdata and having > it overridden by the framework. There is nothing wrong in using > dev_get_drvdata(), assuming the correct type is known and here it's > guaranteed to be v4l2_subdev for the v4l2-subdev framework. > > In fact i2c_get_clientdata() [1] is just a wrapper that calls > dev_get_drvdata(&client->dev). And funny future would be that if we don't do this now we will have plenty of janitors to DoS attack us with the fix in the future. > [1] https://elixir.bootlin.com/linux/v5.9-rc3/source/include/linux/i2c.h#L351 -- With Best Regards, Andy Shevchenko
