1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 * relocate_kernel.S - put the kernel image in place to boot
4 * Copyright (C) 2002-2005 Eric Biederman <ebiederm@xmission.com>
5 */
6
7#include <linux/linkage.h>
8#include <asm/page_types.h>
9#include <asm/kexec.h>
10#include <asm/processor-flags.h>
11#include <asm/pgtable_types.h>
12#include <asm/nospec-branch.h>
13#include <asm/unwind_hints.h>
14
15/*
16 * Must be relocatable PIC code callable as a C function, in particular
17 * there must be a plain RET and not jump to return thunk.
18 */
19
20#define PTR(x) (x << 3)
21#define PAGE_ATTR (_PAGE_PRESENT | _PAGE_RW | _PAGE_ACCESSED | _PAGE_DIRTY)
22
23/*
24 * control_page + KEXEC_CONTROL_CODE_MAX_SIZE
25 * ~ control_page + PAGE_SIZE are used as data storage and stack for
26 * jumping back
27 */
28#define DATA(offset) (KEXEC_CONTROL_CODE_MAX_SIZE+(offset))
29
30/* Minimal CPU state */
31#define RSP DATA(0x0)
32#define CR0 DATA(0x8)
33#define CR3 DATA(0x10)
34#define CR4 DATA(0x18)
35
36/* other data */
37#define CP_PA_TABLE_PAGE DATA(0x20)
38#define CP_PA_SWAP_PAGE DATA(0x28)
39#define CP_PA_BACKUP_PAGES_MAP DATA(0x30)
40
41 .text
42 .align PAGE_SIZE
43 .code64
44SYM_CODE_START_NOALIGN(relocate_range)
45SYM_CODE_START_NOALIGN(relocate_kernel)
46 UNWIND_HINT_END_OF_STACK
47 ANNOTATE_NOENDBR
48 /*
49 * %rdi indirection_page
50 * %rsi page_list
51 * %rdx start address
52 * %rcx preserve_context
53 * %r8 host_mem_enc_active
54 */
55
56 /* Save the CPU context, used for jumping back */
57 pushq %rbx
58 pushq %rbp
59 pushq %r12
60 pushq %r13
61 pushq %r14
62 pushq %r15
63 pushf
64
65 movq PTR(VA_CONTROL_PAGE)(%rsi), %r11
66 movq %rsp, RSP(%r11)
67 movq %cr0, %rax
68 movq %rax, CR0(%r11)
69 movq %cr3, %rax
70 movq %rax, CR3(%r11)
71 movq %cr4, %rax
72 movq %rax, CR4(%r11)
73
74 /* Save CR4. Required to enable the right paging mode later. */
75 movq %rax, %r13
76
77 /* zero out flags, and disable interrupts */
78 pushq $0
79 popfq
80
81 /* Save SME active flag */
82 movq %r8, %r12
83
84 /*
85 * get physical address of control page now
86 * this is impossible after page table switch
87 */
88 movq PTR(PA_CONTROL_PAGE)(%rsi), %r8
89
90 /* get physical address of page table now too */
91 movq PTR(PA_TABLE_PAGE)(%rsi), %r9
92
93 /* get physical address of swap page now */
94 movq PTR(PA_SWAP_PAGE)(%rsi), %r10
95
96 /* save some information for jumping back */
97 movq %r9, CP_PA_TABLE_PAGE(%r11)
98 movq %r10, CP_PA_SWAP_PAGE(%r11)
99 movq %rdi, CP_PA_BACKUP_PAGES_MAP(%r11)
100
101 /* Switch to the identity mapped page tables */
102 movq %r9, %cr3
103
104 /* setup a new stack at the end of the physical control page */
105 lea PAGE_SIZE(%r8), %rsp
106
107 /* jump to identity mapped page */
108 addq $(identity_mapped - relocate_kernel), %r8
109 pushq %r8
110 ANNOTATE_UNRET_SAFE
111 ret
112 int3
113SYM_CODE_END(relocate_kernel)
114
115SYM_CODE_START_LOCAL_NOALIGN(identity_mapped)
116 UNWIND_HINT_END_OF_STACK
117 /* set return address to 0 if not preserving context */
118 pushq $0
119 /* store the start address on the stack */
120 pushq %rdx
121
122 /*
123 * Clear X86_CR4_CET (if it was set) such that we can clear CR0_WP
124 * below.
125 */
126 movq %cr4, %rax
127 andq $~(X86_CR4_CET), %rax
128 movq %rax, %cr4
129
130 /*
131 * Set cr0 to a known state:
132 * - Paging enabled
133 * - Alignment check disabled
134 * - Write protect disabled
135 * - No task switch
136 * - Don't do FP software emulation.
137 * - Protected mode enabled
138 */
139 movq %cr0, %rax
140 andq $~(X86_CR0_AM | X86_CR0_WP | X86_CR0_TS | X86_CR0_EM), %rax
141 orl $(X86_CR0_PG | X86_CR0_PE), %eax
142 movq %rax, %cr0
143
144 /*
145 * Set cr4 to a known state:
146 * - physical address extension enabled
147 * - 5-level paging, if it was enabled before
148 */
149 movl $X86_CR4_PAE, %eax
150 testq $X86_CR4_LA57, %r13
151 jz 1f
152 orl $X86_CR4_LA57, %eax
1531:
154 movq %rax, %cr4
155
156 jmp 1f
1571:
158
159 /* Flush the TLB (needed?) */
160 movq %r9, %cr3
161
162 /*
163 * If SME is active, there could be old encrypted cache line
164 * entries that will conflict with the now unencrypted memory
165 * used by kexec. Flush the caches before copying the kernel.
166 */
167 testq %r12, %r12
168 jz 1f
169 wbinvd
1701:
171
172 movq %rcx, %r11
173 call swap_pages
174
175 /*
176 * To be certain of avoiding problems with self-modifying code
177 * I need to execute a serializing instruction here.
178 * So I flush the TLB by reloading %cr3 here, it's handy,
179 * and not processor dependent.
180 */
181 movq %cr3, %rax
182 movq %rax, %cr3
183
184 /*
185 * set all of the registers to known values
186 * leave %rsp alone
187 */
188
189 testq %r11, %r11
190 jnz 1f
191 xorl %eax, %eax
192 xorl %ebx, %ebx
193 xorl %ecx, %ecx
194 xorl %edx, %edx
195 xorl %esi, %esi
196 xorl %edi, %edi
197 xorl %ebp, %ebp
198 xorl %r8d, %r8d
199 xorl %r9d, %r9d
200 xorl %r10d, %r10d
201 xorl %r11d, %r11d
202 xorl %r12d, %r12d
203 xorl %r13d, %r13d
204 xorl %r14d, %r14d
205 xorl %r15d, %r15d
206
207 ANNOTATE_UNRET_SAFE
208 ret
209 int3
210
2111:
212 popq %rdx
213 leaq PAGE_SIZE(%r10), %rsp
214 ANNOTATE_RETPOLINE_SAFE
215 call *%rdx
216
217 /* get the re-entry point of the peer system */
218 movq 0(%rsp), %rbp
219 leaq relocate_kernel(%rip), %r8
220 movq CP_PA_SWAP_PAGE(%r8), %r10
221 movq CP_PA_BACKUP_PAGES_MAP(%r8), %rdi
222 movq CP_PA_TABLE_PAGE(%r8), %rax
223 movq %rax, %cr3
224 lea PAGE_SIZE(%r8), %rsp
225 call swap_pages
226 movq $virtual_mapped, %rax
227 pushq %rax
228 ANNOTATE_UNRET_SAFE
229 ret
230 int3
231SYM_CODE_END(identity_mapped)
232
233SYM_CODE_START_LOCAL_NOALIGN(virtual_mapped)
234 UNWIND_HINT_END_OF_STACK
235 ANNOTATE_NOENDBR // RET target, above
236 movq RSP(%r8), %rsp
237 movq CR4(%r8), %rax
238 movq %rax, %cr4
239 movq CR3(%r8), %rax
240 movq CR0(%r8), %r8
241 movq %rax, %cr3
242 movq %r8, %cr0
243 movq %rbp, %rax
244
245 popf
246 popq %r15
247 popq %r14
248 popq %r13
249 popq %r12
250 popq %rbp
251 popq %rbx
252 ANNOTATE_UNRET_SAFE
253 ret
254 int3
255SYM_CODE_END(virtual_mapped)
256
257 /* Do the copies */
258SYM_CODE_START_LOCAL_NOALIGN(swap_pages)
259 UNWIND_HINT_END_OF_STACK
260 movq %rdi, %rcx /* Put the page_list in %rcx */
261 xorl %edi, %edi
262 xorl %esi, %esi
263 jmp 1f
264
2650: /* top, read another word for the indirection page */
266
267 movq (%rbx), %rcx
268 addq $8, %rbx
2691:
270 testb $0x1, %cl /* is it a destination page? */
271 jz 2f
272 movq %rcx, %rdi
273 andq $0xfffffffffffff000, %rdi
274 jmp 0b
2752:
276 testb $0x2, %cl /* is it an indirection page? */
277 jz 2f
278 movq %rcx, %rbx
279 andq $0xfffffffffffff000, %rbx
280 jmp 0b
2812:
282 testb $0x4, %cl /* is it the done indicator? */
283 jz 2f
284 jmp 3f
2852:
286 testb $0x8, %cl /* is it the source indicator? */
287 jz 0b /* Ignore it otherwise */
288 movq %rcx, %rsi /* For ever source page do a copy */
289 andq $0xfffffffffffff000, %rsi
290
291 movq %rdi, %rdx
292 movq %rsi, %rax
293
294 movq %r10, %rdi
295 movl $512, %ecx
296 rep ; movsq
297
298 movq %rax, %rdi
299 movq %rdx, %rsi
300 movl $512, %ecx
301 rep ; movsq
302
303 movq %rdx, %rdi
304 movq %r10, %rsi
305 movl $512, %ecx
306 rep ; movsq
307
308 lea PAGE_SIZE(%rax), %rsi
309 jmp 0b
3103:
311 ANNOTATE_UNRET_SAFE
312 ret
313 int3
314SYM_CODE_END(swap_pages)
315
316 .skip KEXEC_CONTROL_CODE_MAX_SIZE - (. - relocate_kernel), 0xcc
317SYM_CODE_END(relocate_range);
318

source code of linux/arch/x86/kernel/relocate_kernel_64.S