On Sun, 2024-02-11 at 18:55 +0200, Andy Shevchenko wrote: > On Sat, Feb 10, 2024 at 10:57 PM Nuno Sa via B4 Relay > <devnull+nuno.sa.analog.com@xxxxxxxxxx> wrote: > > > Changes in v11: > > - Patch 6 > > * Directly use dev in devm_iio_backend_get(); > > * Move comment above the proper place. > > - Patch 7 > > * Added blank line between includes (to logically separate them); > > * Move back to 10 millisecond sleep; > > * Constify expected_ver and removed unneeded cast. > > Assuming that the fwnode vs. device reference count is fine (as we > expect the backends to call respective APIs and hence have backend > device pointer valid despote fwnode reference being dropped before > getting the device) and timeout thingy had been reverted to the > original value, FWIW, > Reviewed-by: Andy Shevchenko <andy.shevchenko@xxxxxxxxx> > for patches 4-7 (OF is not my area of interest at all :). > Thanks for the ack! If I understood the above correctly, we should be fine. Unless the fwnode refcount can reach 0 without unbinding the device from it's driver (which I don't think it's possible). Note the backend will be freed on the unbind path and through devlinks, we guarantee that all the consumers are unbound before us... - Nuno Sá >
