xref: /linux/arch/x86/kernel/reboot.c (revision 6e7fd890f1d6ac83805409e9c346240de2705584)
1 // SPDX-License-Identifier: GPL-2.0
2 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
3 
4 #include <linux/export.h>
5 #include <linux/reboot.h>
6 #include <linux/init.h>
7 #include <linux/pm.h>
8 #include <linux/efi.h>
9 #include <linux/dmi.h>
10 #include <linux/sched.h>
11 #include <linux/tboot.h>
12 #include <linux/delay.h>
13 #include <linux/objtool.h>
14 #include <linux/pgtable.h>
15 #include <linux/kexec.h>
16 #include <acpi/reboot.h>
17 #include <asm/io.h>
18 #include <asm/apic.h>
19 #include <asm/io_apic.h>
20 #include <asm/desc.h>
21 #include <asm/hpet.h>
22 #include <asm/proto.h>
23 #include <asm/reboot_fixups.h>
24 #include <asm/reboot.h>
25 #include <asm/pci_x86.h>
26 #include <asm/cpu.h>
27 #include <asm/nmi.h>
28 #include <asm/smp.h>
29 
30 #include <linux/ctype.h>
31 #include <linux/mc146818rtc.h>
32 #include <asm/realmode.h>
33 #include <asm/x86_init.h>
34 #include <asm/efi.h>
35 
36 /*
37  * Power off function, if any
38  */
39 void (*pm_power_off)(void);
40 EXPORT_SYMBOL(pm_power_off);
41 
42 /*
43  * This is set if we need to go through the 'emergency' path.
44  * When machine_emergency_restart() is called, we may be on
45  * an inconsistent state and won't be able to do a clean cleanup
46  */
47 static int reboot_emergency;
48 
49 /* This is set by the PCI code if either type 1 or type 2 PCI is detected */
50 bool port_cf9_safe = false;
51 
52 /*
53  * Reboot options and system auto-detection code provided by
54  * Dell Inc. so their systems "just work". :-)
55  */
56 
57 /*
58  * Some machines require the "reboot=a" commandline options
59  */
60 static int __init set_acpi_reboot(const struct dmi_system_id *d)
61 {
62 	if (reboot_type != BOOT_ACPI) {
63 		reboot_type = BOOT_ACPI;
64 		pr_info("%s series board detected. Selecting %s-method for reboots.\n",
65 			d->ident, "ACPI");
66 	}
67 	return 0;
68 }
69 
70 /*
71  * Some machines require the "reboot=b" or "reboot=k"  commandline options,
72  * this quirk makes that automatic.
73  */
74 static int __init set_bios_reboot(const struct dmi_system_id *d)
75 {
76 	if (reboot_type != BOOT_BIOS) {
77 		reboot_type = BOOT_BIOS;
78 		pr_info("%s series board detected. Selecting %s-method for reboots.\n",
79 			d->ident, "BIOS");
80 	}
81 	return 0;
82 }
83 
84 /*
85  * Some machines don't handle the default ACPI reboot method and
86  * require the EFI reboot method:
87  */
88 static int __init set_efi_reboot(const struct dmi_system_id *d)
89 {
90 	if (reboot_type != BOOT_EFI && !efi_runtime_disabled()) {
91 		reboot_type = BOOT_EFI;
92 		pr_info("%s series board detected. Selecting EFI-method for reboot.\n", d->ident);
93 	}
94 	return 0;
95 }
96 
97 void __noreturn machine_real_restart(unsigned int type)
98 {
99 	local_irq_disable();
100 
101 	/*
102 	 * Write zero to CMOS register number 0x0f, which the BIOS POST
103 	 * routine will recognize as telling it to do a proper reboot.  (Well
104 	 * that's what this book in front of me says -- it may only apply to
105 	 * the Phoenix BIOS though, it's not clear).  At the same time,
106 	 * disable NMIs by setting the top bit in the CMOS address register,
107 	 * as we're about to do peculiar things to the CPU.  I'm not sure if
108 	 * `outb_p' is needed instead of just `outb'.  Use it to be on the
109 	 * safe side.  (Yes, CMOS_WRITE does outb_p's. -  Paul G.)
110 	 */
111 	spin_lock(&rtc_lock);
112 	CMOS_WRITE(0x00, 0x8f);
113 	spin_unlock(&rtc_lock);
114 
115 	/*
116 	 * Switch to the trampoline page table.
117 	 */
118 	load_trampoline_pgtable();
119 
120 	/* Jump to the identity-mapped low memory code */
121 #ifdef CONFIG_X86_32
122 	asm volatile("jmpl *%0" : :
123 		     "rm" (real_mode_header->machine_real_restart_asm),
124 		     "a" (type));
125 #else
126 	asm volatile("ljmpl *%0" : :
127 		     "m" (real_mode_header->machine_real_restart_asm),
128 		     "D" (type));
129 #endif
130 	unreachable();
131 }
132 #ifdef CONFIG_APM_MODULE
133 EXPORT_SYMBOL(machine_real_restart);
134 #endif
135 STACK_FRAME_NON_STANDARD(machine_real_restart);
136 
137 /*
138  * Some Apple MacBook and MacBookPro's needs reboot=p to be able to reboot
139  */
140 static int __init set_pci_reboot(const struct dmi_system_id *d)
141 {
142 	if (reboot_type != BOOT_CF9_FORCE) {
143 		reboot_type = BOOT_CF9_FORCE;
144 		pr_info("%s series board detected. Selecting %s-method for reboots.\n",
145 			d->ident, "PCI");
146 	}
147 	return 0;
148 }
149 
150 static int __init set_kbd_reboot(const struct dmi_system_id *d)
151 {
152 	if (reboot_type != BOOT_KBD) {
153 		reboot_type = BOOT_KBD;
154 		pr_info("%s series board detected. Selecting %s-method for reboot.\n",
155 			d->ident, "KBD");
156 	}
157 	return 0;
158 }
159 
160 /*
161  * This is a single dmi_table handling all reboot quirks.
162  */
163 static const struct dmi_system_id reboot_dmi_table[] __initconst = {
164 
165 	/* Acer */
166 	{	/* Handle reboot issue on Acer Aspire one */
167 		.callback = set_kbd_reboot,
168 		.ident = "Acer Aspire One A110",
169 		.matches = {
170 			DMI_MATCH(DMI_SYS_VENDOR, "Acer"),
171 			DMI_MATCH(DMI_PRODUCT_NAME, "AOA110"),
172 		},
173 	},
174 	{	/* Handle reboot issue on Acer TravelMate X514-51T */
175 		.callback = set_efi_reboot,
176 		.ident = "Acer TravelMate X514-51T",
177 		.matches = {
178 			DMI_MATCH(DMI_SYS_VENDOR, "Acer"),
179 			DMI_MATCH(DMI_PRODUCT_NAME, "TravelMate X514-51T"),
180 		},
181 	},
182 
183 	/* Apple */
184 	{	/* Handle problems with rebooting on Apple MacBook5 */
185 		.callback = set_pci_reboot,
186 		.ident = "Apple MacBook5",
187 		.matches = {
188 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
189 			DMI_MATCH(DMI_PRODUCT_NAME, "MacBook5"),
190 		},
191 	},
192 	{	/* Handle problems with rebooting on Apple MacBook6,1 */
193 		.callback = set_pci_reboot,
194 		.ident = "Apple MacBook6,1",
195 		.matches = {
196 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
197 			DMI_MATCH(DMI_PRODUCT_NAME, "MacBook6,1"),
198 		},
199 	},
200 	{	/* Handle problems with rebooting on Apple MacBookPro5 */
201 		.callback = set_pci_reboot,
202 		.ident = "Apple MacBookPro5",
203 		.matches = {
204 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
205 			DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5"),
206 		},
207 	},
208 	{	/* Handle problems with rebooting on Apple Macmini3,1 */
209 		.callback = set_pci_reboot,
210 		.ident = "Apple Macmini3,1",
211 		.matches = {
212 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
213 			DMI_MATCH(DMI_PRODUCT_NAME, "Macmini3,1"),
214 		},
215 	},
216 	{	/* Handle problems with rebooting on the iMac9,1. */
217 		.callback = set_pci_reboot,
218 		.ident = "Apple iMac9,1",
219 		.matches = {
220 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
221 			DMI_MATCH(DMI_PRODUCT_NAME, "iMac9,1"),
222 		},
223 	},
224 	{	/* Handle problems with rebooting on the iMac10,1. */
225 		.callback = set_pci_reboot,
226 		.ident = "Apple iMac10,1",
227 		.matches = {
228 		    DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
229 		    DMI_MATCH(DMI_PRODUCT_NAME, "iMac10,1"),
230 		},
231 	},
232 
233 	/* ASRock */
234 	{	/* Handle problems with rebooting on ASRock Q1900DC-ITX */
235 		.callback = set_pci_reboot,
236 		.ident = "ASRock Q1900DC-ITX",
237 		.matches = {
238 			DMI_MATCH(DMI_BOARD_VENDOR, "ASRock"),
239 			DMI_MATCH(DMI_BOARD_NAME, "Q1900DC-ITX"),
240 		},
241 	},
242 
243 	/* ASUS */
244 	{	/* Handle problems with rebooting on ASUS P4S800 */
245 		.callback = set_bios_reboot,
246 		.ident = "ASUS P4S800",
247 		.matches = {
248 			DMI_MATCH(DMI_BOARD_VENDOR, "ASUSTeK Computer INC."),
249 			DMI_MATCH(DMI_BOARD_NAME, "P4S800"),
250 		},
251 	},
252 	{	/* Handle problems with rebooting on ASUS EeeBook X205TA */
253 		.callback = set_acpi_reboot,
254 		.ident = "ASUS EeeBook X205TA",
255 		.matches = {
256 			DMI_MATCH(DMI_SYS_VENDOR, "ASUSTeK COMPUTER INC."),
257 			DMI_MATCH(DMI_PRODUCT_NAME, "X205TA"),
258 		},
259 	},
260 	{	/* Handle problems with rebooting on ASUS EeeBook X205TAW */
261 		.callback = set_acpi_reboot,
262 		.ident = "ASUS EeeBook X205TAW",
263 		.matches = {
264 			DMI_MATCH(DMI_SYS_VENDOR, "ASUSTeK COMPUTER INC."),
265 			DMI_MATCH(DMI_PRODUCT_NAME, "X205TAW"),
266 		},
267 	},
268 
269 	/* Certec */
270 	{       /* Handle problems with rebooting on Certec BPC600 */
271 		.callback = set_pci_reboot,
272 		.ident = "Certec BPC600",
273 		.matches = {
274 			DMI_MATCH(DMI_SYS_VENDOR, "Certec"),
275 			DMI_MATCH(DMI_PRODUCT_NAME, "BPC600"),
276 		},
277 	},
278 
279 	/* Dell */
280 	{	/* Handle problems with rebooting on Dell DXP061 */
281 		.callback = set_bios_reboot,
282 		.ident = "Dell DXP061",
283 		.matches = {
284 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
285 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell DXP061"),
286 		},
287 	},
288 	{	/* Handle problems with rebooting on Dell E520's */
289 		.callback = set_bios_reboot,
290 		.ident = "Dell E520",
291 		.matches = {
292 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
293 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell DM061"),
294 		},
295 	},
296 	{	/* Handle problems with rebooting on the Latitude E5410. */
297 		.callback = set_pci_reboot,
298 		.ident = "Dell Latitude E5410",
299 		.matches = {
300 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
301 			DMI_MATCH(DMI_PRODUCT_NAME, "Latitude E5410"),
302 		},
303 	},
304 	{	/* Handle problems with rebooting on the Latitude E5420. */
305 		.callback = set_pci_reboot,
306 		.ident = "Dell Latitude E5420",
307 		.matches = {
308 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
309 			DMI_MATCH(DMI_PRODUCT_NAME, "Latitude E5420"),
310 		},
311 	},
312 	{	/* Handle problems with rebooting on the Latitude E6320. */
313 		.callback = set_pci_reboot,
314 		.ident = "Dell Latitude E6320",
315 		.matches = {
316 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
317 			DMI_MATCH(DMI_PRODUCT_NAME, "Latitude E6320"),
318 		},
319 	},
320 	{	/* Handle problems with rebooting on the Latitude E6420. */
321 		.callback = set_pci_reboot,
322 		.ident = "Dell Latitude E6420",
323 		.matches = {
324 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
325 			DMI_MATCH(DMI_PRODUCT_NAME, "Latitude E6420"),
326 		},
327 	},
328 	{	/* Handle problems with rebooting on Dell Optiplex 330 with 0KP561 */
329 		.callback = set_bios_reboot,
330 		.ident = "Dell OptiPlex 330",
331 		.matches = {
332 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
333 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 330"),
334 			DMI_MATCH(DMI_BOARD_NAME, "0KP561"),
335 		},
336 	},
337 	{	/* Handle problems with rebooting on Dell Optiplex 360 with 0T656F */
338 		.callback = set_bios_reboot,
339 		.ident = "Dell OptiPlex 360",
340 		.matches = {
341 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
342 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 360"),
343 			DMI_MATCH(DMI_BOARD_NAME, "0T656F"),
344 		},
345 	},
346 	{	/* Handle problems with rebooting on Dell Optiplex 745's SFF */
347 		.callback = set_bios_reboot,
348 		.ident = "Dell OptiPlex 745",
349 		.matches = {
350 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
351 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
352 		},
353 	},
354 	{	/* Handle problems with rebooting on Dell Optiplex 745's DFF */
355 		.callback = set_bios_reboot,
356 		.ident = "Dell OptiPlex 745",
357 		.matches = {
358 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
359 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
360 			DMI_MATCH(DMI_BOARD_NAME, "0MM599"),
361 		},
362 	},
363 	{	/* Handle problems with rebooting on Dell Optiplex 745 with 0KW626 */
364 		.callback = set_bios_reboot,
365 		.ident = "Dell OptiPlex 745",
366 		.matches = {
367 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
368 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
369 			DMI_MATCH(DMI_BOARD_NAME, "0KW626"),
370 		},
371 	},
372 	{	/* Handle problems with rebooting on Dell OptiPlex 760 with 0G919G */
373 		.callback = set_bios_reboot,
374 		.ident = "Dell OptiPlex 760",
375 		.matches = {
376 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
377 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 760"),
378 			DMI_MATCH(DMI_BOARD_NAME, "0G919G"),
379 		},
380 	},
381 	{	/* Handle problems with rebooting on the OptiPlex 990. */
382 		.callback = set_pci_reboot,
383 		.ident = "Dell OptiPlex 990 BIOS A0x",
384 		.matches = {
385 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
386 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 990"),
387 			DMI_MATCH(DMI_BIOS_VERSION, "A0"),
388 		},
389 	},
390 	{	/* Handle problems with rebooting on Dell 300's */
391 		.callback = set_bios_reboot,
392 		.ident = "Dell PowerEdge 300",
393 		.matches = {
394 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
395 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 300/"),
396 		},
397 	},
398 	{	/* Handle problems with rebooting on Dell 1300's */
399 		.callback = set_bios_reboot,
400 		.ident = "Dell PowerEdge 1300",
401 		.matches = {
402 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
403 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 1300/"),
404 		},
405 	},
406 	{	/* Handle problems with rebooting on Dell 2400's */
407 		.callback = set_bios_reboot,
408 		.ident = "Dell PowerEdge 2400",
409 		.matches = {
410 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
411 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"),
412 		},
413 	},
414 	{	/* Handle problems with rebooting on the Dell PowerEdge C6100. */
415 		.callback = set_pci_reboot,
416 		.ident = "Dell PowerEdge C6100",
417 		.matches = {
418 			DMI_MATCH(DMI_SYS_VENDOR, "Dell"),
419 			DMI_MATCH(DMI_PRODUCT_NAME, "C6100"),
420 		},
421 	},
422 	{	/* Handle problems with rebooting on the Precision M6600. */
423 		.callback = set_pci_reboot,
424 		.ident = "Dell Precision M6600",
425 		.matches = {
426 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
427 			DMI_MATCH(DMI_PRODUCT_NAME, "Precision M6600"),
428 		},
429 	},
430 	{	/* Handle problems with rebooting on Dell T5400's */
431 		.callback = set_bios_reboot,
432 		.ident = "Dell Precision T5400",
433 		.matches = {
434 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
435 			DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"),
436 		},
437 	},
438 	{	/* Handle problems with rebooting on Dell T7400's */
439 		.callback = set_bios_reboot,
440 		.ident = "Dell Precision T7400",
441 		.matches = {
442 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
443 			DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T7400"),
444 		},
445 	},
446 	{	/* Handle problems with rebooting on Dell XPS710 */
447 		.callback = set_bios_reboot,
448 		.ident = "Dell XPS710",
449 		.matches = {
450 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
451 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
452 		},
453 	},
454 	{	/* Handle problems with rebooting on Dell Optiplex 7450 AIO */
455 		.callback = set_acpi_reboot,
456 		.ident = "Dell OptiPlex 7450 AIO",
457 		.matches = {
458 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
459 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 7450 AIO"),
460 		},
461 	},
462 
463 	/* Hewlett-Packard */
464 	{	/* Handle problems with rebooting on HP laptops */
465 		.callback = set_bios_reboot,
466 		.ident = "HP Compaq Laptop",
467 		.matches = {
468 			DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
469 			DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq"),
470 		},
471 	},
472 
473 	{	/* PCIe Wifi card isn't detected after reboot otherwise */
474 		.callback = set_pci_reboot,
475 		.ident = "Zotac ZBOX CI327 nano",
476 		.matches = {
477 			DMI_MATCH(DMI_SYS_VENDOR, "NA"),
478 			DMI_MATCH(DMI_PRODUCT_NAME, "ZBOX-CI327NANO-GS-01"),
479 		},
480 	},
481 
482 	/* Sony */
483 	{	/* Handle problems with rebooting on Sony VGN-Z540N */
484 		.callback = set_bios_reboot,
485 		.ident = "Sony VGN-Z540N",
486 		.matches = {
487 			DMI_MATCH(DMI_SYS_VENDOR, "Sony Corporation"),
488 			DMI_MATCH(DMI_PRODUCT_NAME, "VGN-Z540N"),
489 		},
490 	},
491 
492 	{ }
493 };
494 
495 static int __init reboot_init(void)
496 {
497 	int rv;
498 
499 	/*
500 	 * Only do the DMI check if reboot_type hasn't been overridden
501 	 * on the command line
502 	 */
503 	if (!reboot_default)
504 		return 0;
505 
506 	/*
507 	 * The DMI quirks table takes precedence. If no quirks entry
508 	 * matches and the ACPI Hardware Reduced bit is set and EFI
509 	 * runtime services are enabled, force EFI reboot.
510 	 */
511 	rv = dmi_check_system(reboot_dmi_table);
512 
513 	if (!rv && efi_reboot_required() && !efi_runtime_disabled())
514 		reboot_type = BOOT_EFI;
515 
516 	return 0;
517 }
518 core_initcall(reboot_init);
519 
520 static inline void kb_wait(void)
521 {
522 	int i;
523 
524 	for (i = 0; i < 0x10000; i++) {
525 		if ((inb(0x64) & 0x02) == 0)
526 			break;
527 		udelay(2);
528 	}
529 }
530 
531 static inline void nmi_shootdown_cpus_on_restart(void);
532 
533 #if IS_ENABLED(CONFIG_KVM_INTEL) || IS_ENABLED(CONFIG_KVM_AMD)
534 /* RCU-protected callback to disable virtualization prior to reboot. */
535 static cpu_emergency_virt_cb __rcu *cpu_emergency_virt_callback;
536 
537 void cpu_emergency_register_virt_callback(cpu_emergency_virt_cb *callback)
538 {
539 	if (WARN_ON_ONCE(rcu_access_pointer(cpu_emergency_virt_callback)))
540 		return;
541 
542 	rcu_assign_pointer(cpu_emergency_virt_callback, callback);
543 }
544 EXPORT_SYMBOL_GPL(cpu_emergency_register_virt_callback);
545 
546 void cpu_emergency_unregister_virt_callback(cpu_emergency_virt_cb *callback)
547 {
548 	if (WARN_ON_ONCE(rcu_access_pointer(cpu_emergency_virt_callback) != callback))
549 		return;
550 
551 	rcu_assign_pointer(cpu_emergency_virt_callback, NULL);
552 	synchronize_rcu();
553 }
554 EXPORT_SYMBOL_GPL(cpu_emergency_unregister_virt_callback);
555 
556 /*
557  * Disable virtualization, i.e. VMX or SVM, to ensure INIT is recognized during
558  * reboot.  VMX blocks INIT if the CPU is post-VMXON, and SVM blocks INIT if
559  * GIF=0, i.e. if the crash occurred between CLGI and STGI.
560  */
561 void cpu_emergency_disable_virtualization(void)
562 {
563 	cpu_emergency_virt_cb *callback;
564 
565 	/*
566 	 * IRQs must be disabled as KVM enables virtualization in hardware via
567 	 * function call IPIs, i.e. IRQs need to be disabled to guarantee
568 	 * virtualization stays disabled.
569 	 */
570 	lockdep_assert_irqs_disabled();
571 
572 	rcu_read_lock();
573 	callback = rcu_dereference(cpu_emergency_virt_callback);
574 	if (callback)
575 		callback();
576 	rcu_read_unlock();
577 }
578 
579 static void emergency_reboot_disable_virtualization(void)
580 {
581 	local_irq_disable();
582 
583 	/*
584 	 * Disable virtualization on all CPUs before rebooting to avoid hanging
585 	 * the system, as VMX and SVM block INIT when running in the host.
586 	 *
587 	 * We can't take any locks and we may be on an inconsistent state, so
588 	 * use NMIs as IPIs to tell the other CPUs to disable VMX/SVM and halt.
589 	 *
590 	 * Do the NMI shootdown even if virtualization is off on _this_ CPU, as
591 	 * other CPUs may have virtualization enabled.
592 	 */
593 	if (rcu_access_pointer(cpu_emergency_virt_callback)) {
594 		/* Safely force _this_ CPU out of VMX/SVM operation. */
595 		cpu_emergency_disable_virtualization();
596 
597 		/* Disable VMX/SVM and halt on other CPUs. */
598 		nmi_shootdown_cpus_on_restart();
599 	}
600 }
601 #else
602 static void emergency_reboot_disable_virtualization(void) { }
603 #endif /* CONFIG_KVM_INTEL || CONFIG_KVM_AMD */
604 
605 void __attribute__((weak)) mach_reboot_fixups(void)
606 {
607 }
608 
609 /*
610  * To the best of our knowledge Windows compatible x86 hardware expects
611  * the following on reboot:
612  *
613  * 1) If the FADT has the ACPI reboot register flag set, try it
614  * 2) If still alive, write to the keyboard controller
615  * 3) If still alive, write to the ACPI reboot register again
616  * 4) If still alive, write to the keyboard controller again
617  * 5) If still alive, call the EFI runtime service to reboot
618  * 6) If no EFI runtime service, call the BIOS to do a reboot
619  *
620  * We default to following the same pattern. We also have
621  * two other reboot methods: 'triple fault' and 'PCI', which
622  * can be triggered via the reboot= kernel boot option or
623  * via quirks.
624  *
625  * This means that this function can never return, it can misbehave
626  * by not rebooting properly and hanging.
627  */
628 static void native_machine_emergency_restart(void)
629 {
630 	int i;
631 	int attempt = 0;
632 	int orig_reboot_type = reboot_type;
633 	unsigned short mode;
634 
635 	if (reboot_emergency)
636 		emergency_reboot_disable_virtualization();
637 
638 	tboot_shutdown(TB_SHUTDOWN_REBOOT);
639 
640 	/* Tell the BIOS if we want cold or warm reboot */
641 	mode = reboot_mode == REBOOT_WARM ? 0x1234 : 0;
642 	*((unsigned short *)__va(0x472)) = mode;
643 
644 	/*
645 	 * If an EFI capsule has been registered with the firmware then
646 	 * override the reboot= parameter.
647 	 */
648 	if (efi_capsule_pending(NULL)) {
649 		pr_info("EFI capsule is pending, forcing EFI reboot.\n");
650 		reboot_type = BOOT_EFI;
651 	}
652 
653 	for (;;) {
654 		/* Could also try the reset bit in the Hammer NB */
655 		switch (reboot_type) {
656 		case BOOT_ACPI:
657 			acpi_reboot();
658 			reboot_type = BOOT_KBD;
659 			break;
660 
661 		case BOOT_KBD:
662 			mach_reboot_fixups(); /* For board specific fixups */
663 
664 			for (i = 0; i < 10; i++) {
665 				kb_wait();
666 				udelay(50);
667 				outb(0xfe, 0x64); /* Pulse reset low */
668 				udelay(50);
669 			}
670 			if (attempt == 0 && orig_reboot_type == BOOT_ACPI) {
671 				attempt = 1;
672 				reboot_type = BOOT_ACPI;
673 			} else {
674 				reboot_type = BOOT_EFI;
675 			}
676 			break;
677 
678 		case BOOT_EFI:
679 			efi_reboot(reboot_mode, NULL);
680 			reboot_type = BOOT_BIOS;
681 			break;
682 
683 		case BOOT_BIOS:
684 			machine_real_restart(MRR_BIOS);
685 
686 			/* We're probably dead after this, but... */
687 			reboot_type = BOOT_CF9_SAFE;
688 			break;
689 
690 		case BOOT_CF9_FORCE:
691 			port_cf9_safe = true;
692 			fallthrough;
693 
694 		case BOOT_CF9_SAFE:
695 			if (port_cf9_safe) {
696 				u8 reboot_code = reboot_mode == REBOOT_WARM ?  0x06 : 0x0E;
697 				u8 cf9 = inb(0xcf9) & ~reboot_code;
698 				outb(cf9|2, 0xcf9); /* Request hard reset */
699 				udelay(50);
700 				/* Actually do the reset */
701 				outb(cf9|reboot_code, 0xcf9);
702 				udelay(50);
703 			}
704 			reboot_type = BOOT_TRIPLE;
705 			break;
706 
707 		case BOOT_TRIPLE:
708 			idt_invalidate();
709 			__asm__ __volatile__("int3");
710 
711 			/* We're probably dead after this, but... */
712 			reboot_type = BOOT_KBD;
713 			break;
714 		}
715 	}
716 }
717 
718 void native_machine_shutdown(void)
719 {
720 	/*
721 	 * Call enc_kexec_begin() while all CPUs are still active and
722 	 * interrupts are enabled. This will allow all in-flight memory
723 	 * conversions to finish cleanly.
724 	 */
725 	if (kexec_in_progress)
726 		x86_platform.guest.enc_kexec_begin();
727 
728 	/* Stop the cpus and apics */
729 #ifdef CONFIG_X86_IO_APIC
730 	/*
731 	 * Disabling IO APIC before local APIC is a workaround for
732 	 * erratum AVR31 in "Intel Atom Processor C2000 Product Family
733 	 * Specification Update". In this situation, interrupts that target
734 	 * a Logical Processor whose Local APIC is either in the process of
735 	 * being hardware disabled or software disabled are neither delivered
736 	 * nor discarded. When this erratum occurs, the processor may hang.
737 	 *
738 	 * Even without the erratum, it still makes sense to quiet IO APIC
739 	 * before disabling Local APIC.
740 	 */
741 	clear_IO_APIC();
742 #endif
743 
744 #ifdef CONFIG_SMP
745 	/*
746 	 * Stop all of the others. Also disable the local irq to
747 	 * not receive the per-cpu timer interrupt which may trigger
748 	 * scheduler's load balance.
749 	 */
750 	local_irq_disable();
751 	stop_other_cpus();
752 #endif
753 
754 	lapic_shutdown();
755 	restore_boot_irq_mode();
756 
757 #ifdef CONFIG_HPET_TIMER
758 	hpet_disable();
759 #endif
760 
761 #ifdef CONFIG_X86_64
762 	x86_platform.iommu_shutdown();
763 #endif
764 
765 	if (kexec_in_progress)
766 		x86_platform.guest.enc_kexec_finish();
767 }
768 
769 static void __machine_emergency_restart(int emergency)
770 {
771 	reboot_emergency = emergency;
772 	machine_ops.emergency_restart();
773 }
774 
775 static void native_machine_restart(char *__unused)
776 {
777 	pr_notice("machine restart\n");
778 
779 	if (!reboot_force)
780 		machine_shutdown();
781 	__machine_emergency_restart(0);
782 }
783 
784 static void native_machine_halt(void)
785 {
786 	/* Stop other cpus and apics */
787 	machine_shutdown();
788 
789 	tboot_shutdown(TB_SHUTDOWN_HALT);
790 
791 	stop_this_cpu(NULL);
792 }
793 
794 static void native_machine_power_off(void)
795 {
796 	if (kernel_can_power_off()) {
797 		if (!reboot_force)
798 			machine_shutdown();
799 		do_kernel_power_off();
800 	}
801 	/* A fallback in case there is no PM info available */
802 	tboot_shutdown(TB_SHUTDOWN_HALT);
803 }
804 
805 struct machine_ops machine_ops __ro_after_init = {
806 	.power_off = native_machine_power_off,
807 	.shutdown = native_machine_shutdown,
808 	.emergency_restart = native_machine_emergency_restart,
809 	.restart = native_machine_restart,
810 	.halt = native_machine_halt,
811 #ifdef CONFIG_CRASH_DUMP
812 	.crash_shutdown = native_machine_crash_shutdown,
813 #endif
814 };
815 
816 void machine_power_off(void)
817 {
818 	machine_ops.power_off();
819 }
820 
821 void machine_shutdown(void)
822 {
823 	machine_ops.shutdown();
824 }
825 
826 void machine_emergency_restart(void)
827 {
828 	__machine_emergency_restart(1);
829 }
830 
831 void machine_restart(char *cmd)
832 {
833 	machine_ops.restart(cmd);
834 }
835 
836 void machine_halt(void)
837 {
838 	machine_ops.halt();
839 }
840 
841 #ifdef CONFIG_CRASH_DUMP
842 void machine_crash_shutdown(struct pt_regs *regs)
843 {
844 	machine_ops.crash_shutdown(regs);
845 }
846 #endif
847 
848 /* This is the CPU performing the emergency shutdown work. */
849 int crashing_cpu = -1;
850 
851 #if defined(CONFIG_SMP)
852 
853 static nmi_shootdown_cb shootdown_callback;
854 
855 static atomic_t waiting_for_crash_ipi;
856 static int crash_ipi_issued;
857 
858 static int crash_nmi_callback(unsigned int val, struct pt_regs *regs)
859 {
860 	int cpu;
861 
862 	cpu = raw_smp_processor_id();
863 
864 	/*
865 	 * Don't do anything if this handler is invoked on crashing cpu.
866 	 * Otherwise, system will completely hang. Crashing cpu can get
867 	 * an NMI if system was initially booted with nmi_watchdog parameter.
868 	 */
869 	if (cpu == crashing_cpu)
870 		return NMI_HANDLED;
871 	local_irq_disable();
872 
873 	if (shootdown_callback)
874 		shootdown_callback(cpu, regs);
875 
876 	/*
877 	 * Prepare the CPU for reboot _after_ invoking the callback so that the
878 	 * callback can safely use virtualization instructions, e.g. VMCLEAR.
879 	 */
880 	cpu_emergency_disable_virtualization();
881 
882 	atomic_dec(&waiting_for_crash_ipi);
883 
884 	if (smp_ops.stop_this_cpu) {
885 		smp_ops.stop_this_cpu();
886 		unreachable();
887 	}
888 
889 	/* Assume hlt works */
890 	halt();
891 	for (;;)
892 		cpu_relax();
893 
894 	return NMI_HANDLED;
895 }
896 
897 /**
898  * nmi_shootdown_cpus - Stop other CPUs via NMI
899  * @callback:	Optional callback to be invoked from the NMI handler
900  *
901  * The NMI handler on the remote CPUs invokes @callback, if not
902  * NULL, first and then disables virtualization to ensure that
903  * INIT is recognized during reboot.
904  *
905  * nmi_shootdown_cpus() can only be invoked once. After the first
906  * invocation all other CPUs are stuck in crash_nmi_callback() and
907  * cannot respond to a second NMI.
908  */
909 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
910 {
911 	unsigned long msecs;
912 
913 	local_irq_disable();
914 
915 	/*
916 	 * Avoid certain doom if a shootdown already occurred; re-registering
917 	 * the NMI handler will cause list corruption, modifying the callback
918 	 * will do who knows what, etc...
919 	 */
920 	if (WARN_ON_ONCE(crash_ipi_issued))
921 		return;
922 
923 	/* Make a note of crashing cpu. Will be used in NMI callback. */
924 	crashing_cpu = safe_smp_processor_id();
925 
926 	shootdown_callback = callback;
927 
928 	atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
929 	/* Would it be better to replace the trap vector here? */
930 	if (register_nmi_handler(NMI_LOCAL, crash_nmi_callback,
931 				 NMI_FLAG_FIRST, "crash"))
932 		return;		/* Return what? */
933 	/*
934 	 * Ensure the new callback function is set before sending
935 	 * out the NMI
936 	 */
937 	wmb();
938 
939 	apic_send_IPI_allbutself(NMI_VECTOR);
940 
941 	/* Kick CPUs looping in NMI context. */
942 	WRITE_ONCE(crash_ipi_issued, 1);
943 
944 	msecs = 1000; /* Wait at most a second for the other cpus to stop */
945 	while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
946 		mdelay(1);
947 		msecs--;
948 	}
949 
950 	/*
951 	 * Leave the nmi callback set, shootdown is a one-time thing.  Clearing
952 	 * the callback could result in a NULL pointer dereference if a CPU
953 	 * (finally) responds after the timeout expires.
954 	 */
955 }
956 
957 static inline void nmi_shootdown_cpus_on_restart(void)
958 {
959 	if (!crash_ipi_issued)
960 		nmi_shootdown_cpus(NULL);
961 }
962 
963 /*
964  * Check if the crash dumping IPI got issued and if so, call its callback
965  * directly. This function is used when we have already been in NMI handler.
966  * It doesn't return.
967  */
968 void run_crash_ipi_callback(struct pt_regs *regs)
969 {
970 	if (crash_ipi_issued)
971 		crash_nmi_callback(0, regs);
972 }
973 
974 /* Override the weak function in kernel/panic.c */
975 void __noreturn nmi_panic_self_stop(struct pt_regs *regs)
976 {
977 	while (1) {
978 		/* If no CPU is preparing crash dump, we simply loop here. */
979 		run_crash_ipi_callback(regs);
980 		cpu_relax();
981 	}
982 }
983 
984 #else /* !CONFIG_SMP */
985 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
986 {
987 	/* No other CPUs to shoot down */
988 }
989 
990 static inline void nmi_shootdown_cpus_on_restart(void) { }
991 
992 void run_crash_ipi_callback(struct pt_regs *regs)
993 {
994 }
995 #endif
996