On Wed, Feb 12, 2020 at 04:57:19PM -0500, Arvind Sankar wrote: > However, I'm also getting confused about how mixed-mode works at all if > we have more than 4Gb RAM, which it seems we want to support as we take > pains to allocate a stack below 4Gb for the thunk. But in this case, the > kernel text and stack could be above 4Gb physically, so rather than > using a 1:1 map we'd need to find some space in the low 4Gb of virtual > addresses to map those, but I don't see where we do this? Also, that > dynamically allocated variable_name could be above 4G as well. Verified in QEMU that mixed mode crashes if physical RAM can extend above 4G. Are there any such real devices in existence? Should we check that and disable runtime services if so?
