On Thu, 11 Jun 2020 15:02:05 +0200 Eugene Zaikonnikov <ez@xxxxxxxxxxxxx> wrote: > Jonathan Cameron <Jonathan.Cameron@xxxxxxxxxx> writes: > > > On Mon, 18 May 2020 19:15:27 +0000 > > Eugene Zalkonnikov <ez@xxxxxxxxxxxxx> wrote: > > > > Hi Eugene > > > >> HDC2010/2080 driver and sysfs documentation for its heater element. > >> > > > > Would be nice to have a little more info on the device for anyone reading > > the log after we commit this. > > > Ok I'll expand the description. > > > There is a potential double free in the remove path due to mix of > > devm_iio_device_register and iio_device_unregister. > > > Right, overlooked this. Will changine the unregister to the managed > version suffice? Whilst that would work it would be a bad idea. You need the unmanaged unregister before disabling auto measurement (to remove the userspace interfaces before changing things so they won't work). So just use unmanaged iio_device_register in probe (and reorder to put the unregister before the sensor stop). I think that will also mirror the ordering in probe which should always be the case unless there is a very good reason to do otherwise. Jonathan > > > Also, the error paths in probe don't undo the various bits of setup > > that it feels like they should. This might not matter in reality > > but the code will be more obviously correct if it is fine to reset > > the amm bit. > > > I'll add that. >
