illumos gate

Yet another bug where someone casts a pointer to a larger sized integer and risks sign extension!
Are we having fun yet?

In a struct mdnm_params, as generally given to metaioctl, we have a devname field that's conceptually a pointer but is actually uint64_t rather than uintptr_t. As such, when 32bit, there's a risk the actual pointer assigned to it will be sign-extended, it needs to be cast to uintptr_t first to make sure that doesn't occur.

I'm not going to get into the rathole of why this is a fixed 64bit type rather than something the size of a native pointer, I can see how we'd get here, and changing it is prohibitively costly in test/time.