Possible uninitialised value for map->memory causes failure to map range
When I refactored the x86 method in 95fbfeea I introduced a couple of bugs in the x86 backend, as far as I can tell:
- There should be no probe() call in create()
- There should be no mapping of regions during probe()
I have provided a patch series in !13 (merged) that addresses these bugs as well as some refactoring to make it neater with hurd usage. Please review my changes.