[mirror_edk2.git] / UefiCpuPkg / CpuMpPei / X64 / MpFuncs.asm

;------------------------------------------------------------------------------ ;\r
; Copyright (c) 2015, Intel Corporation. All rights reserved.<BR>\r
; This program and the accompanying materials\r
; are licensed and made available under the terms and conditions of the BSD License\r
; which accompanies this distribution.  The full text of the license may be found at\r
; http://opensource.org/licenses/bsd-license.php.\r
;\r
; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,\r
; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.\r
;\r
; Module Name:\r
;\r
;   MpFuncs32.asm\r
;\r
; Abstract:\r
;\r
;   This is the assembly code for MP support\r
;\r
;-------------------------------------------------------------------------------\r
\r
include  MpEqu.inc\r
extern   InitializeFloatingPointUnits:PROC\r
\r
.code\r
;-------------------------------------------------------------------------------------\r
;RendezvousFunnelProc  procedure follows. All APs execute their procedure. This\r
;procedure serializes all the AP processors through an Init sequence. It must be\r
;noted that APs arrive here very raw...ie: real mode, no stack.\r
;ALSO THIS PROCEDURE IS EXECUTED BY APs ONLY ON 16 BIT MODE. HENCE THIS PROC\r
;IS IN MACHINE CODE.\r
;-------------------------------------------------------------------------------------\r
RendezvousFunnelProc   PROC  PUBLIC\r
RendezvousFunnelProcStart::\r
; At this point CS = 0x(vv00) and ip= 0x0.\r
; Save BIST information to ebp firstly\r
    db 66h,  08bh, 0e8h               ; mov        ebp, eax    ; save BIST information\r
\r
    db 8ch,0c8h                       ; mov        ax,cs\r
    db 8eh,0d8h                       ; mov        ds,ax\r
    db 8eh,0c0h                       ; mov        es,ax\r
    db 8eh,0d0h                       ; mov        ss,ax\r
    db 33h,0c0h                       ; xor        ax,ax\r
    db 8eh,0e0h                       ; mov        fs,ax\r
    db 8eh,0e8h                       ; mov        gs,ax\r
\r
    db 0BEh                           ; opcode of mov si, mem16\r
    dw BufferStartLocation            ; mov        si, BufferStartLocation\r
    db 66h,  8Bh, 1Ch                 ; mov        ebx,dword ptr [si]\r
\r
    db 0BFh                           ; opcode of mov di, mem16\r
    dw PmodeOffsetLocation            ; mov        di, PmodeOffsetLocation\r
    db 66h,  8Bh, 05h                 ; mov        eax,dword ptr [di]\r
    db 8Bh,  0F8h                     ; mov        di, ax\r
    db 83h,  0EFh,06h                 ; sub        di, 06h\r
    db 66h,  03h, 0C3h                ; add        eax, ebx\r
    db 66h,  89h, 05h                 ; mov        dword ptr [di],eax\r
\r
    db 0BFh                           ; opcode of mov di, mem16\r
    dw LmodeOffsetLocation            ; mov        di, LmodeOffsetLocation\r
    db 66h,  8Bh, 05h                 ; mov        eax,dword ptr [di]\r
    db 8Bh,  0F8h                     ; mov        di, ax\r
    db 83h,  0EFh,06h                 ; sub        di, 06h\r
    db 66h,  03h, 0C3h                ; add        eax, ebx\r
    db 66h,  89h, 05h                 ; mov        dword ptr [di],eax\r
\r
    db 0BEh\r
    dw Cr3Location                    ; mov        si, Cr3Location\r
    db 66h,  8Bh, 0Ch                 ; mov        ecx,dword ptr [si]     ; ECX is keeping the value of CR3\r
\r
    db 0BEh                           ; opcode of mov si, mem16\r
    dw GdtrLocation                   ; mov        si, GdtrLocation\r
    db 66h                            ; db         66h\r
    db 2Eh,  0Fh, 01h, 14h            ; lgdt       fword ptr cs:[si]\r
\r
    db 0BEh\r
    dw IdtrLocation                   ; mov        si, IdtrLocation\r
    db 66h                            ; db         66h\r
    db 2Eh,0Fh, 01h, 1Ch              ; lidt       fword ptr cs:[si]\r
\r
    db 33h,  0C0h                     ; xor        ax,  ax\r
    db 8Eh,  0D8h                     ; mov        ds,  ax\r
\r
    db 0Fh,  20h, 0C0h                ; mov        eax, cr0               ;Get control register 0\r
    db 66h,  83h, 0C8h, 03h           ; or         eax, 000000003h        ;Set PE bit (bit #0) & MP\r
    db 0Fh,  22h, 0C0h                ; mov        cr0, eax\r
\r
    db 66h,  67h, 0EAh                ; far jump\r
    dd 0h                             ; 32-bit offset\r
    dw PROTECT_MODE_CS                ; 16-bit selector\r
\r
Flat32Start::                         ; protected mode entry point\r
    mov        ax, PROTECT_MODE_DS\r
    mov        ds, ax\r
    mov        es, ax\r
    mov        fs, ax\r
    mov        gs, ax\r
    mov        ss, ax\r
\r
    db 0Fh,  20h,  0E0h           ; mov        eax, cr4\r
    db 0Fh,  0BAh, 0E8h, 05h      ; bts        eax, 5\r
    db 0Fh,  22h,  0E0h           ; mov        cr4, eax\r
\r
    db 0Fh,  22h,  0D9h           ; mov        cr3, ecx\r
\r
    db 0B9h\r
    dd 0C0000080h                 ; mov        ecx, 0c0000080h     ; EFER MSR number.\r
    db 0Fh,  32h                  ; rdmsr                          ; Read EFER.\r
    db 0Fh,  0BAh, 0E8h, 08h      ; bts        eax, 8              ; Set LME=1.\r
    db 0Fh,  30h                  ; wrmsr                          ; Write EFER.\r
\r
    db 0Fh,  20h,  0C0h           ; mov        eax, cr0            ; Read CR0.\r
    db 0Fh,  0BAh, 0E8h, 1Fh      ; bts        eax, 31             ; Set PG=1.\r
    db 0Fh,  22h,  0C0h           ; mov        cr0, eax            ; Write CR0.\r
\r
LONG_JUMP:\r
    db 67h,  0EAh                 ; far jump\r
    dd 0h                         ; 32-bit offset\r
    dw LONG_MODE_CS               ; 16-bit selector\r
\r
LongModeStart::\r
    mov        ax,  LONG_MODE_DS\r
    mov        ds,  ax\r
    mov        es,  ax\r
    mov        ss,  ax\r
\r
    mov        esi, ebx\r
    mov        edi, esi\r
    add        edi, LockLocation\r
    mov        rax, NotVacantFlag\r
\r
TestLock:\r
    xchg       qword ptr [edi], rax\r
    cmp        rax, NotVacantFlag\r
    jz         TestLock\r
\r
    mov        edi, esi\r
    add        edi, NumApsExecutingLoction\r
    inc        dword ptr [edi]\r
    mov        ebx, dword ptr [edi]\r
\r
ProgramStack:\r
    mov        edi, esi\r
    add        edi, StackSizeLocation\r
    mov        rax, qword ptr [edi]\r
    mov        edi, esi\r
    add        edi, StackStartAddressLocation\r
    add        rax, qword ptr [edi]\r
    mov        rsp, rax\r
    mov        qword ptr [edi], rax\r
\r
Releaselock:\r
    mov        rax, VacantFlag\r
    mov        edi, esi\r
    add        edi, LockLocation\r
    xchg       qword ptr [edi], rax\r
\r
CProcedureInvoke:\r
    push       rbp               ; push BIST data\r
    xor        rbp, rbp          ; clear ebp for call stack trace\r
    push       rbp\r
    mov        rbp, rsp\r
\r
    mov        rax, InitializeFloatingPointUnits\r
    sub        rsp, 20h\r
    call       rax               ; Call assembly function to initialize FPU per UEFI spec\r
    add        rsp, 20h\r
\r
    mov        edx, ebx          ; edx is NumApsExecuting\r
    mov        ecx, esi\r
    add        ecx, LockLocation ; rcx is address of exchange info data buffer\r
\r
    mov        edi, esi\r
    add        edi, ApProcedureLocation\r
    mov        rax, qword ptr [edi]\r
\r
    sub        rsp, 20h\r
    call       rax               ; invoke C function\r
    add        rsp, 20h\r
    jmp        $\r
\r
RendezvousFunnelProc   ENDP\r
RendezvousFunnelProcEnd::\r
\r
AsmCliHltLoop PROC\r
    cli\r
    hlt\r
    jmp $-2\r
AsmCliHltLoop ENDP\r
\r
;-------------------------------------------------------------------------------------\r
;  AsmGetAddressMap (&AddressMap);\r
;-------------------------------------------------------------------------------------\r
AsmGetAddressMap   PROC\r
    mov        rax, offset RendezvousFunnelProcStart\r
    mov        qword ptr [rcx], rax\r
    mov        qword ptr [rcx +  8h], Flat32Start - RendezvousFunnelProcStart\r
    mov        qword ptr [rcx + 10h], LongModeStart - RendezvousFunnelProcStart\r
    mov        qword ptr [rcx + 18h], RendezvousFunnelProcEnd - RendezvousFunnelProcStart\r
    ret\r
AsmGetAddressMap   ENDP\r
\r
;-------------------------------------------------------------------------------------\r
;AsmExchangeRole procedure follows. This procedure executed by current BSP, that is\r
;about to become an AP. It switches it'stack with the current AP.\r
;AsmExchangeRole (IN   CPU_EXCHANGE_INFO    *MyInfo, IN   CPU_EXCHANGE_INFO    *OthersInfo);\r
;-------------------------------------------------------------------------------------\r
AsmExchangeRole   PROC\r
    ; DO NOT call other functions in this function, since 2 CPU may use 1 stack\r
    ; at the same time. If 1 CPU try to call a function, stack will be corrupted.\r
\r
    push       rax\r
    push       rbx\r
    push       rcx\r
    push       rdx\r
    push       rsi\r
    push       rdi\r
    push       rbp\r
    push       r8\r
    push       r9\r
    push       r10\r
    push       r11\r
    push       r12\r
    push       r13\r
    push       r14\r
    push       r15\r
\r
    mov        rax, cr0\r
    push       rax\r
\r
    mov        rax, cr4\r
    push       rax\r
\r
    ; rsi contains MyInfo pointer\r
    mov        rsi, rcx\r
\r
    ; rdi contains OthersInfo pointer\r
    mov        rdi, rdx\r
\r
    ;Store EFLAGS, GDTR and IDTR regiter to stack\r
    pushfq\r
    sgdt       fword ptr [rsi + 16]\r
    sidt       fword ptr [rsi + 26]\r
\r
    ; Store the its StackPointer\r
    mov        qword ptr [rsi + 8], rsp\r
\r
    ; update its switch state to STORED\r
    mov        byte ptr [rsi], CPU_SWITCH_STATE_STORED\r
\r
WaitForOtherStored:\r
    ; wait until the other CPU finish storing its state\r
    cmp        byte ptr [rdi], CPU_SWITCH_STATE_STORED\r
    jz         OtherStored\r
    pause\r
    jmp        WaitForOtherStored\r
\r
OtherStored:\r
    ; Since another CPU already stored its state, load them\r
    ; load GDTR value\r
    lgdt       fword ptr [rdi + 16]\r
\r
    ; load IDTR value\r
    lidt       fword ptr [rdi + 26]\r
\r
    ; load its future StackPointer\r
    mov        rsp, qword ptr [rdi + 8]\r
\r
    ; update the other CPU's switch state to LOADED\r
    mov        byte ptr [rdi], CPU_SWITCH_STATE_LOADED\r
\r
WaitForOtherLoaded:\r
    ; wait until the other CPU finish loading new state,\r
    ; otherwise the data in stack may corrupt\r
    cmp        byte ptr [rsi], CPU_SWITCH_STATE_LOADED\r
    jz         OtherLoaded\r
    pause\r
    jmp        WaitForOtherLoaded\r
\r
OtherLoaded:\r
    ; since the other CPU already get the data it want, leave this procedure\r
    popfq\r
\r
    pop        rax\r
    mov        cr4, rax\r
\r
    pop        rax\r
    mov        cr0, rax\r
\r
    pop        r15\r
    pop        r14\r
    pop        r13\r
    pop        r12\r
    pop        r11\r
    pop        r10\r
    pop        r9\r
    pop        r8\r
    pop        rbp\r
    pop        rdi\r
    pop        rsi\r
    pop        rdx\r
    pop        rcx\r
    pop        rbx\r
    pop        rax\r
\r
    ret\r
AsmExchangeRole   ENDP\r
\r
AsmInitializeGdt   PROC\r
    push       rbp\r
    mov        rbp, rsp\r
\r
    lgdt       fword PTR [rcx]  ; update the GDTR\r
\r
    sub        rsp, 0x10\r
    lea        rax, SetCodeSelectorFarJump\r
    mov        [rsp], rax\r
    mov        rdx, LONG_MODE_CS\r
    mov        [rsp + 4], dx    ; get new CS\r
    jmp        fword ptr [rsp]\r
SetCodeSelectorFarJump:\r
    add        rsp, 0x10\r
\r
    mov        rax, LONG_MODE_DS          ; get new DS\r
    mov        ds, ax\r
    mov        es, ax\r
    mov        fs, ax\r
    mov        gs, ax\r
    mov        ss, ax\r
\r
    pop        rbp\r
    ret\r
AsmInitializeGdt  ENDP\r
\r
END\r
Commit	Line	Data
ea0f431c JF	1	;------------------------------------------------------------------------------ ;\r
	2	; Copyright (c) 2015, Intel Corporation. All rights reserved.<BR>\r
	3	; This program and the accompanying materials\r
	4	; are licensed and made available under the terms and conditions of the BSD License\r
	5	; which accompanies this distribution. The full text of the license may be found at\r
	6	; http://opensource.org/licenses/bsd-license.php.\r
	7	;\r
	8	; THE PROGRAM IS DISTRIBUTED UNDER THE BSD LICENSE ON AN "AS IS" BASIS,\r
	9	; WITHOUT WARRANTIES OR REPRESENTATIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED.\r
	10	;\r
	11	; Module Name:\r
	12	;\r
	13	; MpFuncs32.asm\r
	14	;\r
	15	; Abstract:\r
	16	;\r
	17	; This is the assembly code for MP support\r
	18	;\r
	19	;-------------------------------------------------------------------------------\r
	20	\r
	21	include MpEqu.inc\r
	22	extern InitializeFloatingPointUnits:PROC\r
	23	\r
	24	.code\r
	25	;-------------------------------------------------------------------------------------\r
	26	;RendezvousFunnelProc procedure follows. All APs execute their procedure. This\r
	27	;procedure serializes all the AP processors through an Init sequence. It must be\r
	28	;noted that APs arrive here very raw...ie: real mode, no stack.\r
	29	;ALSO THIS PROCEDURE IS EXECUTED BY APs ONLY ON 16 BIT MODE. HENCE THIS PROC\r
	30	;IS IN MACHINE CODE.\r
	31	;-------------------------------------------------------------------------------------\r
	32	RendezvousFunnelProc PROC PUBLIC\r
	33	RendezvousFunnelProcStart::\r
	34	; At this point CS = 0x(vv00) and ip= 0x0.\r
	35	; Save BIST information to ebp firstly\r
	36	db 66h, 08bh, 0e8h ; mov ebp, eax ; save BIST information\r
	37	\r
	38	db 8ch,0c8h ; mov ax,cs\r
	39	db 8eh,0d8h ; mov ds,ax\r
	40	db 8eh,0c0h ; mov es,ax\r
	41	db 8eh,0d0h ; mov ss,ax\r
	42	db 33h,0c0h ; xor ax,ax\r
	43	db 8eh,0e0h ; mov fs,ax\r
	44	db 8eh,0e8h ; mov gs,ax\r
	45	\r
	46	db 0BEh ; opcode of mov si, mem16\r
	47	dw BufferStartLocation ; mov si, BufferStartLocation\r
	48	db 66h, 8Bh, 1Ch ; mov ebx,dword ptr [si]\r
	49	\r
	50	db 0BFh ; opcode of mov di, mem16\r
	51	dw PmodeOffsetLocation ; mov di, PmodeOffsetLocation\r
	52	db 66h, 8Bh, 05h ; mov eax,dword ptr [di]\r
	53	db 8Bh, 0F8h ; mov di, ax\r
	54	db 83h, 0EFh,06h ; sub di, 06h\r
	55	db 66h, 03h, 0C3h ; add eax, ebx\r
	56	db 66h, 89h, 05h ; mov dword ptr [di],eax\r
	57	\r
	58	db 0BFh ; opcode of mov di, mem16\r
	59	dw LmodeOffsetLocation ; mov di, LmodeOffsetLocation\r
	60	db 66h, 8Bh, 05h ; mov eax,dword ptr [di]\r
	61	db 8Bh, 0F8h ; mov di, ax\r
	62	db 83h, 0EFh,06h ; sub di, 06h\r
	63	db 66h, 03h, 0C3h ; add eax, ebx\r
	64	db 66h, 89h, 05h ; mov dword ptr [di],eax\r
65	\r
66	db 0BEh\r
67	dw Cr3Location ; mov si, Cr3Location\r
68	db 66h, 8Bh, 0Ch ; mov ecx,dword ptr [si] ; ECX is keeping the value of CR3\r
69	\r
70	db 0BEh ; opcode of mov si, mem16\r
71	dw GdtrLocation ; mov si, GdtrLocation\r
72	db 66h ; db 66h\r
73	db 2Eh, 0Fh, 01h, 14h ; lgdt fword ptr cs:[si]\r
74	\r
75	db 0BEh\r
76	dw IdtrLocation ; mov si, IdtrLocation\r
77	db 66h ; db 66h\r
78	db 2Eh,0Fh, 01h, 1Ch ; lidt fword ptr cs:[si]\r
79	\r
80	db 33h, 0C0h ; xor ax, ax\r
81	db 8Eh, 0D8h ; mov ds, ax\r
82	\r
83	db 0Fh, 20h, 0C0h ; mov eax, cr0 ;Get control register 0\r
84	db 66h, 83h, 0C8h, 03h ; or eax, 000000003h ;Set PE bit (bit #0) & MP\r
85	db 0Fh, 22h, 0C0h ; mov cr0, eax\r
86	\r
87	db 66h, 67h, 0EAh ; far jump\r
88	dd 0h ; 32-bit offset\r
89	dw PROTECT_MODE_CS ; 16-bit selector\r
90	\r
91	Flat32Start:: ; protected mode entry point\r
92	mov ax, PROTECT_MODE_DS\r
93	mov ds, ax\r
94	mov es, ax\r
95	mov fs, ax\r
96	mov gs, ax\r
97	mov ss, ax\r
98	\r
99	db 0Fh, 20h, 0E0h ; mov eax, cr4\r
100	db 0Fh, 0BAh, 0E8h, 05h ; bts eax, 5\r
101	db 0Fh, 22h, 0E0h ; mov cr4, eax\r
102	\r
103	db 0Fh, 22h, 0D9h ; mov cr3, ecx\r
104	\r
105	db 0B9h\r
106	dd 0C0000080h ; mov ecx, 0c0000080h ; EFER MSR number.\r
107	db 0Fh, 32h ; rdmsr ; Read EFER.\r
108	db 0Fh, 0BAh, 0E8h, 08h ; bts eax, 8 ; Set LME=1.\r
109	db 0Fh, 30h ; wrmsr ; Write EFER.\r
110	\r
111	db 0Fh, 20h, 0C0h ; mov eax, cr0 ; Read CR0.\r
112	db 0Fh, 0BAh, 0E8h, 1Fh ; bts eax, 31 ; Set PG=1.\r
113	db 0Fh, 22h, 0C0h ; mov cr0, eax ; Write CR0.\r
114	\r
115	LONG_JUMP:\r
116	db 67h, 0EAh ; far jump\r
117	dd 0h ; 32-bit offset\r
118	dw LONG_MODE_CS ; 16-bit selector\r
119	\r
120	LongModeStart::\r
121	mov ax, LONG_MODE_DS\r
122	mov ds, ax\r
123	mov es, ax\r
124	mov ss, ax\r
125	\r
126	mov esi, ebx\r
127	mov edi, esi\r
128	add edi, LockLocation\r
129	mov rax, NotVacantFlag\r
130	\r
131	TestLock:\r
132	xchg qword ptr [edi], rax\r
133	cmp rax, NotVacantFlag\r
134	jz TestLock\r
135	\r
136	mov edi, esi\r
137	add edi, NumApsExecutingLoction\r
138	inc dword ptr [edi]\r
139	mov ebx, dword ptr [edi]\r
140	\r
141	ProgramStack:\r
142	mov edi, esi\r
143	add edi, StackSizeLocation\r
144	mov rax, qword ptr [edi]\r
145	mov edi, esi\r
146	add edi, StackStartAddressLocation\r
147	add rax, qword ptr [edi]\r
148	mov rsp, rax\r
149	mov qword ptr [edi], rax\r
150	\r
151	Releaselock:\r
152	mov rax, VacantFlag\r
153	mov edi, esi\r
154	add edi, LockLocation\r
155	xchg qword ptr [edi], rax\r
156	\r
157	CProcedureInvoke:\r
158	push rbp ; push BIST data\r
159	xor rbp, rbp ; clear ebp for call stack trace\r
160	push rbp\r
161	mov rbp, rsp\r
162	\r
163	mov rax, InitializeFloatingPointUnits\r
164	sub rsp, 20h\r
165	call rax ; Call assembly function to initialize FPU per UEFI spec\r
166	add rsp, 20h\r
167	\r
168	mov edx, ebx ; edx is NumApsExecuting\r
169	mov ecx, esi\r
170	add ecx, LockLocation ; rcx is address of exchange info data buffer\r
171	\r
172	mov edi, esi\r
173	add edi, ApProcedureLocation\r
174	mov rax, qword ptr [edi]\r
175	\r
176	sub rsp, 20h\r
177	call rax ; invoke C function\r
178	add rsp, 20h\r
179	jmp $\r
180	\r
181	RendezvousFunnelProc ENDP\r
182	RendezvousFunnelProcEnd::\r
183	\r
184	AsmCliHltLoop PROC\r
185	cli\r
186	hlt\r
187	jmp $-2\r
188	AsmCliHltLoop ENDP\r
189	\r
190	;-------------------------------------------------------------------------------------\r
191	; AsmGetAddressMap (&AddressMap);\r
192	;-------------------------------------------------------------------------------------\r
193	AsmGetAddressMap PROC\r
194	mov rax, offset RendezvousFunnelProcStart\r
195	mov qword ptr [rcx], rax\r
196	mov qword ptr [rcx + 8h], Flat32Start - RendezvousFunnelProcStart\r
197	mov qword ptr [rcx + 10h], LongModeStart - RendezvousFunnelProcStart\r
198	mov qword ptr [rcx + 18h], RendezvousFunnelProcEnd - RendezvousFunnelProcStart\r
199	ret\r
200	AsmGetAddressMap ENDP\r
201	\r
202	;-------------------------------------------------------------------------------------\r
203	;AsmExchangeRole procedure follows. This procedure executed by current BSP, that is\r
204	;about to become an AP. It switches it'stack with the current AP.\r
205	;AsmExchangeRole (IN CPU_EXCHANGE_INFO MyInfo, IN CPU_EXCHANGE_INFO OthersInfo);\r
206	;-------------------------------------------------------------------------------------\r
207	AsmExchangeRole PROC\r
208	; DO NOT call other functions in this function, since 2 CPU may use 1 stack\r
209	; at the same time. If 1 CPU try to call a function, stack will be corrupted.\r
210	\r
211	push rax\r
212	push rbx\r
213	push rcx\r
214	push rdx\r
215	push rsi\r
216	push rdi\r
217	push rbp\r
218	push r8\r
219	push r9\r
220	push r10\r
221	push r11\r
222	push r12\r
223	push r13\r
224	push r14\r
225	push r15\r
226	\r
227	mov rax, cr0\r
228	push rax\r
229	\r
230	mov rax, cr4\r
231	push rax\r
232	\r
233	; rsi contains MyInfo pointer\r
234	mov rsi, rcx\r
235	\r
236	; rdi contains OthersInfo pointer\r
237	mov rdi, rdx\r
238	\r
239	;Store EFLAGS, GDTR and IDTR regiter to stack\r
240	pushfq\r
241	sgdt fword ptr [rsi + 16]\r
242	sidt fword ptr [rsi + 26]\r
243	\r
244	; Store the its StackPointer\r
245	mov qword ptr [rsi + 8], rsp\r
246	\r
247	; update its switch state to STORED\r
248	mov byte ptr [rsi], CPU_SWITCH_STATE_STORED\r
249	\r
250	WaitForOtherStored:\r
251	; wait until the other CPU finish storing its state\r
252	cmp byte ptr [rdi], CPU_SWITCH_STATE_STORED\r
253	jz OtherStored\r
254	pause\r
255	jmp WaitForOtherStored\r
256	\r
257	OtherStored:\r
258	; Since another CPU already stored its state, load them\r
259	; load GDTR value\r
260	lgdt fword ptr [rdi + 16]\r
261	\r
262	; load IDTR value\r
263	lidt fword ptr [rdi + 26]\r
264	\r
265	; load its future StackPointer\r
266	mov rsp, qword ptr [rdi + 8]\r
267	\r
268	; update the other CPU's switch state to LOADED\r
269	mov byte ptr [rdi], CPU_SWITCH_STATE_LOADED\r
270	\r
271	WaitForOtherLoaded:\r
272	; wait until the other CPU finish loading new state,\r
273	; otherwise the data in stack may corrupt\r
274	cmp byte ptr [rsi], CPU_SWITCH_STATE_LOADED\r
275	jz OtherLoaded\r
276	pause\r
277	jmp WaitForOtherLoaded\r
278	\r
279	OtherLoaded:\r
280	; since the other CPU already get the data it want, leave this procedure\r
281	popfq\r
282	\r
283	pop rax\r
284	mov cr4, rax\r
285	\r
286	pop rax\r
287	mov cr0, rax\r
288	\r
289	pop r15\r
290	pop r14\r
291	pop r13\r
292	pop r12\r
293	pop r11\r
294	pop r10\r
295	pop r9\r
296	pop r8\r
297	pop rbp\r
298	pop rdi\r
299	pop rsi\r
300	pop rdx\r
301	pop rcx\r
302	pop rbx\r
303	pop rax\r
304	\r
305	ret\r
306	AsmExchangeRole ENDP\r
307	\r
308	AsmInitializeGdt PROC\r
309	push rbp\r
310	mov rbp, rsp\r
311	\r
312	lgdt fword PTR [rcx] ; update the GDTR\r
313	\r
314	sub rsp, 0x10\r
315	lea rax, SetCodeSelectorFarJump\r
316	mov [rsp], rax\r
317	mov rdx, LONG_MODE_CS\r
318	mov [rsp + 4], dx ; get new CS\r
319	jmp fword ptr [rsp]\r
320	SetCodeSelectorFarJump:\r
321	add rsp, 0x10\r
322	\r
323	mov rax, LONG_MODE_DS ; get new DS\r
324	mov ds, ax\r
325	mov es, ax\r
326	mov fs, ax\r
327	mov gs, ax\r
328	mov ss, ax\r
329	\r
330	pop rbp\r
331	ret\r
332	AsmInitializeGdt ENDP\r
333	\r
334	END\r