Reviewing this code in more detail... I think clearing the upper 32-bits is a bug. These functions are supposed to return pointers, and since this is X64 code those pointers could be anywhere in address space. The fact that the FSP is in XIP NEM, which on current Intel platforms just happens to be mapped <4GB does not mean that this pointer will always be 4GB. Therefore, I believe the correct course of action is to delete the AND/MOV instructions in question.
Thanks, Nate -----Original Message----- From: Pedro Falcato <pedro.falc...@gmail.com> Sent: Monday, March 20, 2023 12:31 AM To: devel@edk2.groups.io; Ni, Ray <ray...@intel.com> Cc: Chiu, Chasel <chasel.c...@intel.com>; Desimone, Nathaniel L <nathaniel.l.desim...@intel.com>; Zeng, Star <star.z...@intel.com> Subject: Re: [edk2-devel] [PATCH v2] IntelFsp2Pkg: Fix NASM X64 build warnings. On Mon, Mar 20, 2023 at 6:03 AM Ni, Ray <ray...@intel.com> wrote: > > > ASM_PFX(FspInfoHeaderRelativeOff): > > > > DD 0x12345678 ; This value must be patched by the > > build script > > > > - and rax, 0xffffffff > > > > + mov eax, eax ; equal to and rax, 0xFFFFFFFF > > Based on the discussion, we know "mov eax, eax" clears upper 32bits of RAX. > But this code looks very confusing. Is there any other instruction that can > do the same thing? Hi Ray, Any instruction that writes to the lower 32-bits should zero the upper bits. (Pardon my AT&T syntax, just reverse the operands for Intel syntax) and $0xffffffff,%eax gets you a 3 byte opcode (since imm8 is signed, you only need 0xff as the immediate) and %eax, %eax gets you 2 bytes mov %eax, %eax gets you 2 bytes even something silly like adding 0 to EAX should work. But in a pure efficiency+size POV, the last 2 instructions should be optimal. -- Pedro -=-=-=-=-=-=-=-=-=-=-=- Groups.io Links: You receive all messages sent to this group. View/Reply Online (#101515): https://edk2.groups.io/g/devel/message/101515 Mute This Topic: https://groups.io/mt/97678369/21656 Group Owner: devel+ow...@edk2.groups.io Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
