Hi Sylwester, On Thu, Sep 20, 2018 at 11:01:26PM +0200, Sylwester Nawrocki wrote: > Hi Sakari, > > On 09/16/2018 12:52 AM, Sakari Ailus wrote: > > v4l2_i2c_subdev_init() sets the name of the sub-devices (as well as > > entities) to what is fairly certainly known to be unique in the system, > > even if there were more devices of the same kind. > > > > These drivers (m5mols, noon010pc30, ov9650, s5c73m3, s5k4ecgx, s5k6aa) set > > the name to the name of the driver or the module while omitting the > > device's I²C address and bus, leaving the devices with a static name and > > effectively limiting the number of such devices in a media device to 1. > > > > Address this by using the name set by the V4L2 framework. > > > > Signed-off-by: Sakari Ailus<sakari.ailus@xxxxxxxxxxxxxxx> > > Reviewed-by: Akinobu Mita<akinobu.mita@xxxxxxxxx> (ov9650) > > I'm not against this patch but please don't expect an ack from me as this > patch breaks existing user space code, scripts using media-ctl, etc. will > need to be updated after kernel upgrade. I'm mostly concerned about ov9650 > as other drivers are likely only used by Samsung or are obsoleted. That is a fair point and also why I originally sent the patch out as RFC... I checked that OV9650 is around 14 years (!) old now, so it's unlikely to appear in modern hardware in dual configurations. I think this patch thus probably causes more trouble than it has chances of fixing anything. I'll submit one that adds a note that the convention is not to be used in new drivers instead --- unless someone strongly feels this patch really should be merged. -- Regards, Sakari Ailus sakari.ailus@xxxxxxxxxxxxxxx
