On Mon Feb 26 20:08:09 2024 +0000, Jacek Caban wrote:
Would it make sense to make sure incoming pPlacedAddress belongs to a
NT mapping? Technically, spec allows that. But more generally in Wine, we can't reliably support applications doing stuff like that. For all we know, such addresses may belong to some Unix libs. Anyway, I left support for allowing applications to use placed maps out of this MR for now.
Do we want to reset windows page protection back to '0' to match
PROT_NONE resulting from VK_MEMORY_UNMAP_RESERVE_BIT_EXT? Why would we want that? I don't see the spec guarantying PROT_NONE.
As to the first point, yeah true I guess if the application is passing in eronious pointers there won't be a good outcome anyway.
As to the second, good point, if we did care, we'd probably have to determine the page protection of the reserve mapping first. As for why we might want that, it was just a general consideration about keeping the win32 prot values in sync with the underlying mappings.