Created attachment 165265 [details]
Suggested fix for the problem (please see comments for details)
While attaching a gpiobus to a GPIO controller with a single pin, the system panics due to a failed KASSERT in gpiobus_init_softc().
This is due to the fact that gpiobus obtains the maximum pin number from the GPIO controller (via GPIO_PIN_MAX) and immediately does a KASSERT, expecting that the obtained value is != 0.
In the case of a GPIO controller which controls a single pin, the returned value will be 0 (0 is the maximum pin number), which is absolutely valid.
Moreover, right after the KASSERT the code in gpiobus_init_softc() actually adjusts the number of pins to the right value.
The suggestion is to change the KASSERT like so:
- KASSERT(sc->sc_npins != 0, ("GPIO device with no pins"));
+ KASSERT(sc->sc_npins >= 0, ("GPIO device with no pins"));
This, I believe, is a better approach than, say, simply postponing the KASSERT until after gpiobus has adjusted the number of pins, since if a broken GPIO controller driver returns a < 0 value we'll be in even more trouble.
I've attached a patch with the suggested fix.
A commit references this bug:
Date: Thu Jan 14 02:58:29 UTC 2016
New revision: 293872
[gpiobus] handle the case of there being a single GPIO pin available.
Submitted by: Stanislav Galabov <email@example.com>