REF: https://bugzilla.tianocore.org/show_bug.cgi?id=1093 Return Stack Buffer (RSB) is used to predict the target of RET instructions. When the RSB underflows, some processors may fall back to using branch predictors. This might impact software using the retpoline mitigation strategy on those processors. This commit will add RSB stuffing logic before returning from SMM (the RSM instruction) to avoid interfering with non-SMM usage of the retpoline technique. After the stuffing, RSB entries will contain a trap like: @SpecTrap: pause lfence jmp @SpecTrap A more detailed explanation of the purpose of commit is under the 'Branch target injection mitigation' section of the below link: https://software.intel.com/security-software-guidance/insights/host-firmware-speculative-execution-side-channel-mitigation Please note that this commit requires further actions (BZ 1091) to remove the duplicated 'StuffRsb.inc' files and merge them into one under a UefiCpuPkg package-level directory (such as UefiCpuPkg/Include/). REF: https://bugzilla.tianocore.org/show_bug.cgi?id=1091 Cc: Jiewen Yao <jiewen.yao@intel.com> Cc: Michael D Kinney <michael.d.kinney@intel.com> Contributed-under: TianoCore Contribution Agreement 1.1 Signed-off-by: Hao Wu <hao.a.wu@intel.com> Acked-by: Laszlo Ersek <lersek@redhat.com> Regression-tested-by: Laszlo Ersek <lersek@redhat.com> Reviewed-by: Eric Dong <eric.dong@intel.com>
153 lines
4.6 KiB
NASM
153 lines
4.6 KiB
NASM
;------------------------------------------------------------------------------ ;
|
|
; Copyright (c) 2016 - 2018, Intel Corporation. All rights reserved.<BR>
|
|
; This program and the accompanying materials
|
|
; are licensed and made available under the terms and conditions of the BSD License
|
|
; which accompanies this distribution. The full text of the license may be found at
|
|
; http://opensource.org/licenses/bsd-license.php.
|
|
;
|
|
; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,
|
|
; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.
|
|
;
|
|
; Module Name:
|
|
;
|
|
; SmmInit.nasm
|
|
;
|
|
; Abstract:
|
|
;
|
|
; Functions for relocating SMBASE's for all processors
|
|
;
|
|
;-------------------------------------------------------------------------------
|
|
|
|
%include "StuffRsb.inc"
|
|
|
|
extern ASM_PFX(SmmInitHandler)
|
|
extern ASM_PFX(mRebasedFlag)
|
|
extern ASM_PFX(mSmmRelocationOriginalAddress)
|
|
|
|
global ASM_PFX(gPatchSmmCr3)
|
|
global ASM_PFX(gPatchSmmCr4)
|
|
global ASM_PFX(gPatchSmmCr0)
|
|
global ASM_PFX(gPatchSmmInitStack)
|
|
global ASM_PFX(gcSmiInitGdtr)
|
|
global ASM_PFX(gcSmmInitSize)
|
|
global ASM_PFX(gcSmmInitTemplate)
|
|
global ASM_PFX(gPatchRebasedFlagAddr32)
|
|
global ASM_PFX(gPatchSmmRelocationOriginalAddressPtr32)
|
|
|
|
%define LONG_MODE_CS 0x38
|
|
|
|
DEFAULT REL
|
|
SECTION .text
|
|
|
|
ASM_PFX(gcSmiInitGdtr):
|
|
DW 0
|
|
DQ 0
|
|
|
|
global ASM_PFX(SmmStartup)
|
|
|
|
BITS 16
|
|
ASM_PFX(SmmStartup):
|
|
mov eax, 0x80000001 ; read capability
|
|
cpuid
|
|
mov ebx, edx ; rdmsr will change edx. keep it in ebx.
|
|
mov eax, strict dword 0 ; source operand will be patched
|
|
ASM_PFX(gPatchSmmCr3):
|
|
mov cr3, eax
|
|
o32 lgdt [cs:ebp + (ASM_PFX(gcSmiInitGdtr) - ASM_PFX(SmmStartup))]
|
|
mov eax, strict dword 0 ; source operand will be patched
|
|
ASM_PFX(gPatchSmmCr4):
|
|
or ah, 2 ; enable XMM registers access
|
|
mov cr4, eax
|
|
mov ecx, 0xc0000080 ; IA32_EFER MSR
|
|
rdmsr
|
|
or ah, BIT0 ; set LME bit
|
|
test ebx, BIT20 ; check NXE capability
|
|
jz .1
|
|
or ah, BIT3 ; set NXE bit
|
|
.1:
|
|
wrmsr
|
|
mov eax, strict dword 0 ; source operand will be patched
|
|
ASM_PFX(gPatchSmmCr0):
|
|
mov cr0, eax ; enable protected mode & paging
|
|
jmp LONG_MODE_CS : dword 0 ; offset will be patched to @LongMode
|
|
@PatchLongModeOffset:
|
|
|
|
BITS 64
|
|
@LongMode: ; long-mode starts here
|
|
mov rsp, strict qword 0 ; source operand will be patched
|
|
ASM_PFX(gPatchSmmInitStack):
|
|
and sp, 0xfff0 ; make sure RSP is 16-byte aligned
|
|
;
|
|
; Accoring to X64 calling convention, XMM0~5 are volatile, we need to save
|
|
; them before calling C-function.
|
|
;
|
|
sub rsp, 0x60
|
|
movdqa [rsp], xmm0
|
|
movdqa [rsp + 0x10], xmm1
|
|
movdqa [rsp + 0x20], xmm2
|
|
movdqa [rsp + 0x30], xmm3
|
|
movdqa [rsp + 0x40], xmm4
|
|
movdqa [rsp + 0x50], xmm5
|
|
|
|
add rsp, -0x20
|
|
call ASM_PFX(SmmInitHandler)
|
|
add rsp, 0x20
|
|
|
|
;
|
|
; Restore XMM0~5 after calling C-function.
|
|
;
|
|
movdqa xmm0, [rsp]
|
|
movdqa xmm1, [rsp + 0x10]
|
|
movdqa xmm2, [rsp + 0x20]
|
|
movdqa xmm3, [rsp + 0x30]
|
|
movdqa xmm4, [rsp + 0x40]
|
|
movdqa xmm5, [rsp + 0x50]
|
|
|
|
StuffRsb64
|
|
rsm
|
|
|
|
BITS 16
|
|
ASM_PFX(gcSmmInitTemplate):
|
|
mov ebp, [cs:@L1 - ASM_PFX(gcSmmInitTemplate) + 0x8000]
|
|
sub ebp, 0x30000
|
|
jmp ebp
|
|
@L1:
|
|
DQ 0; ASM_PFX(SmmStartup)
|
|
|
|
ASM_PFX(gcSmmInitSize): DW $ - ASM_PFX(gcSmmInitTemplate)
|
|
|
|
BITS 64
|
|
global ASM_PFX(SmmRelocationSemaphoreComplete)
|
|
ASM_PFX(SmmRelocationSemaphoreComplete):
|
|
push rax
|
|
mov rax, [ASM_PFX(mRebasedFlag)]
|
|
mov byte [rax], 1
|
|
pop rax
|
|
jmp [ASM_PFX(mSmmRelocationOriginalAddress)]
|
|
|
|
;
|
|
; Semaphore code running in 32-bit mode
|
|
;
|
|
BITS 32
|
|
global ASM_PFX(SmmRelocationSemaphoreComplete32)
|
|
ASM_PFX(SmmRelocationSemaphoreComplete32):
|
|
push eax
|
|
mov eax, strict dword 0 ; source operand will be patched
|
|
ASM_PFX(gPatchRebasedFlagAddr32):
|
|
mov byte [eax], 1
|
|
pop eax
|
|
jmp dword [dword 0] ; destination will be patched
|
|
ASM_PFX(gPatchSmmRelocationOriginalAddressPtr32):
|
|
|
|
BITS 64
|
|
global ASM_PFX(PiSmmCpuSmmInitFixupAddress)
|
|
ASM_PFX(PiSmmCpuSmmInitFixupAddress):
|
|
lea rax, [@LongMode]
|
|
lea rcx, [@PatchLongModeOffset - 6]
|
|
mov dword [rcx], eax
|
|
|
|
lea rax, [ASM_PFX(SmmStartup)]
|
|
lea rcx, [@L1]
|
|
mov qword [rcx], rax
|
|
ret
|