>>> David Woodhouse <dw...@infradead.org> 07/31/17 1:02 PM >>> >On Sun, 2017-07-30 at 00:16 -0600, Jan Beulich wrote: >> > > > David Woodhouse <dw...@infradead.org> 07/20/17 5:22 PM >>> >> > This includes stuff lke the hypercall tables which we really want >> > to be read-only. And they were going into .data.read-mostly. >> Yes, we'd like them to be read-only, but what if EFI properly assigned r/o >> permissions to the .rodata section when loading xen.efi? We'd then be >> unable to apply relocations when switching from 1:1 to virtual mappings >> (see efi_arch_relocate_image()). > > >FWIW it does look like TianoCore has gained the ability to mark >sections as read-only, in January of this year: >https://github.com/tianocore/edk2/commit/d0e92aad46 > >It doesn't actually seem to be complete — even with subsequent fixes >since that commit, it doesn't look like it catches the case of data >sections without EFI_IMAGE_SCN_MEM_WRITE, such as .rodata. > >And even if/when that gets fixed you'll note that the protection is >deliberately torn down in ExitBootServices(), specifically for the case >you're concerned about below — because you'll need to do the >relocations.
As said in an earlier reply, a first pass over relocations is being done long before the call to ExitBootServices(). A minimal adjustment to efi_arch_relocate_image() will be needed anyway, afaict. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
