> On Monday 11 March 2013, Dong Aisheng wrote: > > > > > + dev = driver_find_device(&syscon_driver.driver, NULL, (void *)s, > > > > > + syscon_match_pdevname); > > > > > + if (!dev) > > > > > + return ERR_PTR(-ENODEV); > > > > > > > > Should it be ERR_PTR(-EPROBE_DEFER)? > > > > > > I have no idea what better using here. Think that is not so important > > > since we may have only one possible error code here. > > > > I'm not quite understand your meaning. > > Since the syscon device may be still not registered, > > so it may be better to return a EPROBE_DERFER which is the same > > as dt version. > > I'm guessing that Alexander has not encountered deferred probing yet. > Alexander, the point here is that returning -EPROBE_DEFER has the > advantage that a probe() callback calling this function can > simply return that error code to the driver core. If the driver > core sees -EPROBE_DEFER, it will retry the same probe function > later, after all other device probe functions have been called > and at least one of them was successful. This way you can load the > syscon driver after loading a driver depending on it and everything > will still work.
You are right Arnd. I usually try to control the correct order of loading drivers. Using EPROBE_DEFER in this procedure is more correct. Probably time for v7... I hope it will finally be the last. ---
