On Fri, Dec 10, 2010 at 08:12:53AM +0100, Corentin Chary wrote: > On Fri, Dec 10, 2010 at 1:08 AM, Dmitry Torokhov > <dmitry.torok...@gmail.com> wrote: > > > > .remove is still needed but I guess Corentin refers to using > > platform_driver_probe() instead of platform_driver_register() so the > > megaware_probe can be marked __init and discarded after driver > > initialization has been completed. > > Also, register won't return any error code if the probe function failed, > and the driver will be left in a strange state. > > Calling directly platform_driver_probe() will return an error code. > > But since it's a singleton device, and since the real "probe" code > is already in the module_init function (wmi_has_guid()) call, using > the probe callback to initialize data structures doesn't seems to make > sense. >
What's the other option (provided that we want to keep platform device)? Manual binding device and driver? I think platform_driver_probe() is better tested and is safer, emits all necessary uevents. etc, etc. I'd suggest platform_create_bundle() except I am not quite happy with the API at the moment (needs to also accept drvdata I think). -- Dmitry -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
