On Thu, Dec 26, 2019 at 04:14:07PM +0100, Ard Biesheuvel wrote: > Calling 32-bit EFI runtime services from a 64-bit OS involves > switching back to the flat mapping with a stack carved out of > memory that is 32-bit addressable. > > There is no need to actually execute the 64-bit part of this > routine from the flat mapping as well, as long as the entry > and return address fit in 32 bits. There is also no need to > preserve part of the calling context in global variables: we > can simply preserve the old stack pointer in %r11 across the > call into 32-bit firmware, and use either stack to preserve > other values. > > Signed-off-by: Ard Biesheuvel <ardb@xxxxxxxxxx> > --- > arch/x86/platform/efi/efi_thunk_64.S | 106 ++++++-------------- > 1 file changed, 29 insertions(+), 77 deletions(-) > > diff --git a/arch/x86/platform/efi/efi_thunk_64.S b/arch/x86/platform/efi/efi_thunk_64.S > index 3189f1394701..7357808d3ae8 100644 > --- a/arch/x86/platform/efi/efi_thunk_64.S > +++ b/arch/x86/platform/efi/efi_thunk_64.S > + /* > + * Convert 32-bit status code into 64-bit. > + */ > +1: btrl $31, %eax > + jb 3f > +2: mov %r11, %rsp > pop %rbx > movl %ebx, %ss > pop %rbx > movl %ebx, %es > pop %rbx > movl %ebx, %ds > - > - /* > - * Convert 32-bit status code into 64-bit. > - */ > - test %rax, %rax > - jz 1f > - movl %eax, %ecx > - andl $0x0fffffff, %ecx > - andl $0xf0000000, %eax > - shl $32, %rax > - or %rcx, %rax > -1: Is it worth optimizing the conversion? The entire high nibble is significant according to the spec. It probably doesn't matter except in one potential case: according to the spec, transitioning secure boot status to setup mode by deleting the platform key is allowed to return EFI_WARN_RESET_REQUIRED and AFAICT this can take place after ExitBootServices? Separately, it might be worth considering moving the status translation into C instead of asm for the version in arch/x86/boot/compressed/efi_thunk_64.S -- there are (at least) three protocols that have methods that don't return efi_status_t: DEVICE_PATH_UTILITIES, DEVICE_PATH_TO_TEXT and DEVICE_PATH_FROM_TEXT. If we ever want to use them (eg for debugging), it might be worth having a thunk that doesn't mangle the return value.
