On Mon, 31 Jul 2023 at 14:32, Akihiko Odaki <akihiko.od...@daynix.com> wrote: > > On 2023/07/31 20:44, Peter Maydell wrote: > > On Mon, 31 Jul 2023 at 09:04, Akihiko Odaki <akihiko.od...@daynix.com> > > wrote: > >> > >> MAP_FIXED_NOREPLACE can ensure the mapped address is fixed without > >> concerning that the new mapping overwrites something else. > > > > MAP_FIXED_NOREPLACE only came in with Linux 4.17. So > > I think we still need to handle the "mapped address > > is not the one we asked for" error condition, because > > it can happen on older host kernels that ignore the > > MAP_FIXED_NOREPLACE flag.
> MAP_FIXED_NOREPLACE is substituted with MAP_FIXED before passing to the > host with patch 1. The NOREPLACE constraint is still ensured by > inspecting the guest page table. Oh, I see, this patch is a call to target_mmap(), not host mmap(). Sorry, I misread that. thanks -- PMM
