Hello, [...] > > > I confirmed that the uintptr_t fixed the issue. > > > > We declined a similar fix in the past[1] ... > > > > > I also think that adding a new struct with the mode is overkill. > > > > ... with the hopes that a driver could drop the switch statements in place > > of using the other pattern. Also, to be consistent with other drivers that > > do this already. > > Note that the issue of casting is something we cannot fix easily: > some *_device_id structs use "kernel_ulong_t" for the "data" member, > others use "void *". > > git grep -W "_device_id" -- include/linux/mod_devicetable.h | grep data > > In addition, several drivers use multiple types of device IDs, so you > cannot settle on one type to avoid casts. > > Also, putting enum values in instances of that struct, as suggested, > increases kernel size, for IMHO no additional gain. All good points! Thank you for taking the time to get back to me. Appreciated. :) > If there is more data to put in the struct, I agree it makes sense to use > a struct. Yeah. Perhaps if there is such a need in the future, indeed. Krzysztof
