This patch series makes driver_register() emit an error message and return a failure code instead of triggering a BUG_ON().
The first patch will cause driver_register() to fail gracefully if the driver's bus was not initialized while printing out the name of the faulty driver *and* the name of the involved bus. The second patch records the error code if bus_register() fails. The third and final patch of the series extends the first patch so that the recorded error code is also print out if non-zero. Otherwise, if drv->bus->p is NULL but the error code is zero, then probably bus_register() was never called before. Greg questioned [1] whether [2/3] and [3/3] are necessary: > And really, when has this ever happened? Why would a bus registration > fail and later allow a driver to be registered? I initially assumed that this is what cause me hitting the BUG_ON() which [1/3] replaces: The bus registration failed and then the driver attempts to register itself. But I did not had a chance to verify that. I'll try to do so after my vacation. Meanwhile I hope that at least [1/3] is considered an improvement of the kernel. If so, feel free to pick it up. 1: Message-ID: <[email protected]> Florian Schmaus (3): driver-core: return EINVAL error instead of BUG_ON() driver-core: record error on bus registration driver-core: print bus registration error value drivers/base/bus.c | 9 +++++++-- drivers/base/driver.c | 12 +++++++++++- include/linux/device.h | 1 + 3 files changed, 19 insertions(+), 3 deletions(-) -- 2.16.1
