xref: /linux/arch/x86/kernel/reboot.c (revision dfc349402de8e95f6a42e8341e9ea193b718eee3)
1 #include <linux/module.h>
2 #include <linux/reboot.h>
3 #include <linux/init.h>
4 #include <linux/pm.h>
5 #include <linux/efi.h>
6 #include <linux/dmi.h>
7 #include <linux/sched.h>
8 #include <linux/tboot.h>
9 #include <acpi/reboot.h>
10 #include <asm/io.h>
11 #include <asm/apic.h>
12 #include <asm/desc.h>
13 #include <asm/hpet.h>
14 #include <asm/pgtable.h>
15 #include <asm/proto.h>
16 #include <asm/reboot_fixups.h>
17 #include <asm/reboot.h>
18 #include <asm/pci_x86.h>
19 #include <asm/virtext.h>
20 #include <asm/cpu.h>
21 
22 #ifdef CONFIG_X86_32
23 # include <linux/ctype.h>
24 # include <linux/mc146818rtc.h>
25 #else
26 # include <asm/iommu.h>
27 #endif
28 
29 /*
30  * Power off function, if any
31  */
32 void (*pm_power_off)(void);
33 EXPORT_SYMBOL(pm_power_off);
34 
35 static const struct desc_ptr no_idt = {};
36 static int reboot_mode;
37 enum reboot_type reboot_type = BOOT_KBD;
38 int reboot_force;
39 
40 #if defined(CONFIG_X86_32) && defined(CONFIG_SMP)
41 static int reboot_cpu = -1;
42 #endif
43 
44 /* This is set if we need to go through the 'emergency' path.
45  * When machine_emergency_restart() is called, we may be on
46  * an inconsistent state and won't be able to do a clean cleanup
47  */
48 static int reboot_emergency;
49 
50 /* This is set by the PCI code if either type 1 or type 2 PCI is detected */
51 bool port_cf9_safe = false;
52 
53 /* reboot=b[ios] | s[mp] | t[riple] | k[bd] | e[fi] [, [w]arm | [c]old] | p[ci]
54    warm   Don't set the cold reboot flag
55    cold   Set the cold reboot flag
56    bios   Reboot by jumping through the BIOS (only for X86_32)
57    smp    Reboot by executing reset on BSP or other CPU (only for X86_32)
58    triple Force a triple fault (init)
59    kbd    Use the keyboard controller. cold reset (default)
60    acpi   Use the RESET_REG in the FADT
61    efi    Use efi reset_system runtime service
62    pci    Use the so-called "PCI reset register", CF9
63    force  Avoid anything that could hang.
64  */
65 static int __init reboot_setup(char *str)
66 {
67 	for (;;) {
68 		switch (*str) {
69 		case 'w':
70 			reboot_mode = 0x1234;
71 			break;
72 
73 		case 'c':
74 			reboot_mode = 0;
75 			break;
76 
77 #ifdef CONFIG_X86_32
78 #ifdef CONFIG_SMP
79 		case 's':
80 			if (isdigit(*(str+1))) {
81 				reboot_cpu = (int) (*(str+1) - '0');
82 				if (isdigit(*(str+2)))
83 					reboot_cpu = reboot_cpu*10 + (int)(*(str+2) - '0');
84 			}
85 				/* we will leave sorting out the final value
86 				   when we are ready to reboot, since we might not
87 				   have set up boot_cpu_id or smp_num_cpu */
88 			break;
89 #endif /* CONFIG_SMP */
90 
91 		case 'b':
92 #endif
93 		case 'a':
94 		case 'k':
95 		case 't':
96 		case 'e':
97 		case 'p':
98 			reboot_type = *str;
99 			break;
100 
101 		case 'f':
102 			reboot_force = 1;
103 			break;
104 		}
105 
106 		str = strchr(str, ',');
107 		if (str)
108 			str++;
109 		else
110 			break;
111 	}
112 	return 1;
113 }
114 
115 __setup("reboot=", reboot_setup);
116 
117 
118 #ifdef CONFIG_X86_32
119 /*
120  * Reboot options and system auto-detection code provided by
121  * Dell Inc. so their systems "just work". :-)
122  */
123 
124 /*
125  * Some machines require the "reboot=b"  commandline option,
126  * this quirk makes that automatic.
127  */
128 static int __init set_bios_reboot(const struct dmi_system_id *d)
129 {
130 	if (reboot_type != BOOT_BIOS) {
131 		reboot_type = BOOT_BIOS;
132 		printk(KERN_INFO "%s series board detected. Selecting BIOS-method for reboots.\n", d->ident);
133 	}
134 	return 0;
135 }
136 
137 static struct dmi_system_id __initdata reboot_dmi_table[] = {
138 	{	/* Handle problems with rebooting on Dell E520's */
139 		.callback = set_bios_reboot,
140 		.ident = "Dell E520",
141 		.matches = {
142 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
143 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell DM061"),
144 		},
145 	},
146 	{	/* Handle problems with rebooting on Dell 1300's */
147 		.callback = set_bios_reboot,
148 		.ident = "Dell PowerEdge 1300",
149 		.matches = {
150 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
151 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 1300/"),
152 		},
153 	},
154 	{	/* Handle problems with rebooting on Dell 300's */
155 		.callback = set_bios_reboot,
156 		.ident = "Dell PowerEdge 300",
157 		.matches = {
158 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
159 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 300/"),
160 		},
161 	},
162 	{       /* Handle problems with rebooting on Dell Optiplex 745's SFF*/
163 		.callback = set_bios_reboot,
164 		.ident = "Dell OptiPlex 745",
165 		.matches = {
166 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
167 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
168 		},
169 	},
170 	{       /* Handle problems with rebooting on Dell Optiplex 745's DFF*/
171 		.callback = set_bios_reboot,
172 		.ident = "Dell OptiPlex 745",
173 		.matches = {
174 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
175 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
176 			DMI_MATCH(DMI_BOARD_NAME, "0MM599"),
177 		},
178 	},
179 	{       /* Handle problems with rebooting on Dell Optiplex 745 with 0KW626 */
180 		.callback = set_bios_reboot,
181 		.ident = "Dell OptiPlex 745",
182 		.matches = {
183 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
184 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
185 			DMI_MATCH(DMI_BOARD_NAME, "0KW626"),
186 		},
187 	},
188 	{   /* Handle problems with rebooting on Dell Optiplex 330 with 0KP561 */
189 		.callback = set_bios_reboot,
190 		.ident = "Dell OptiPlex 330",
191 		.matches = {
192 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
193 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 330"),
194 			DMI_MATCH(DMI_BOARD_NAME, "0KP561"),
195 		},
196 	},
197 	{   /* Handle problems with rebooting on Dell Optiplex 360 with 0T656F */
198 		.callback = set_bios_reboot,
199 		.ident = "Dell OptiPlex 360",
200 		.matches = {
201 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
202 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 360"),
203 			DMI_MATCH(DMI_BOARD_NAME, "0T656F"),
204 		},
205 	},
206 	{	/* Handle problems with rebooting on Dell 2400's */
207 		.callback = set_bios_reboot,
208 		.ident = "Dell PowerEdge 2400",
209 		.matches = {
210 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
211 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"),
212 		},
213 	},
214 	{	/* Handle problems with rebooting on Dell T5400's */
215 		.callback = set_bios_reboot,
216 		.ident = "Dell Precision T5400",
217 		.matches = {
218 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
219 			DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"),
220 		},
221 	},
222 	{	/* Handle problems with rebooting on HP laptops */
223 		.callback = set_bios_reboot,
224 		.ident = "HP Compaq Laptop",
225 		.matches = {
226 			DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
227 			DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq"),
228 		},
229 	},
230 	{	/* Handle problems with rebooting on Dell XPS710 */
231 		.callback = set_bios_reboot,
232 		.ident = "Dell XPS710",
233 		.matches = {
234 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
235 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
236 		},
237 	},
238 	{	/* Handle problems with rebooting on Dell DXP061 */
239 		.callback = set_bios_reboot,
240 		.ident = "Dell DXP061",
241 		.matches = {
242 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
243 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell DXP061"),
244 		},
245 	},
246 	{	/* Handle problems with rebooting on Sony VGN-Z540N */
247 		.callback = set_bios_reboot,
248 		.ident = "Sony VGN-Z540N",
249 		.matches = {
250 			DMI_MATCH(DMI_SYS_VENDOR, "Sony Corporation"),
251 			DMI_MATCH(DMI_PRODUCT_NAME, "VGN-Z540N"),
252 		},
253 	},
254 	{	/* Handle problems with rebooting on CompuLab SBC-FITPC2 */
255 		.callback = set_bios_reboot,
256 		.ident = "CompuLab SBC-FITPC2",
257 		.matches = {
258 			DMI_MATCH(DMI_SYS_VENDOR, "CompuLab"),
259 			DMI_MATCH(DMI_PRODUCT_NAME, "SBC-FITPC2"),
260 		},
261 	},
262 	{ }
263 };
264 
265 static int __init reboot_init(void)
266 {
267 	dmi_check_system(reboot_dmi_table);
268 	return 0;
269 }
270 core_initcall(reboot_init);
271 
272 /* The following code and data reboots the machine by switching to real
273    mode and jumping to the BIOS reset entry point, as if the CPU has
274    really been reset.  The previous version asked the keyboard
275    controller to pulse the CPU reset line, which is more thorough, but
276    doesn't work with at least one type of 486 motherboard.  It is easy
277    to stop this code working; hence the copious comments. */
278 static const unsigned long long
279 real_mode_gdt_entries [3] =
280 {
281 	0x0000000000000000ULL,	/* Null descriptor */
282 	0x00009b000000ffffULL,	/* 16-bit real-mode 64k code at 0x00000000 */
283 	0x000093000100ffffULL	/* 16-bit real-mode 64k data at 0x00000100 */
284 };
285 
286 static const struct desc_ptr
287 real_mode_gdt = { sizeof (real_mode_gdt_entries) - 1, (long)real_mode_gdt_entries },
288 real_mode_idt = { 0x3ff, 0 };
289 
290 /* This is 16-bit protected mode code to disable paging and the cache,
291    switch to real mode and jump to the BIOS reset code.
292 
293    The instruction that switches to real mode by writing to CR0 must be
294    followed immediately by a far jump instruction, which set CS to a
295    valid value for real mode, and flushes the prefetch queue to avoid
296    running instructions that have already been decoded in protected
297    mode.
298 
299    Clears all the flags except ET, especially PG (paging), PE
300    (protected-mode enable) and TS (task switch for coprocessor state
301    save).  Flushes the TLB after paging has been disabled.  Sets CD and
302    NW, to disable the cache on a 486, and invalidates the cache.  This
303    is more like the state of a 486 after reset.  I don't know if
304    something else should be done for other chips.
305 
306    More could be done here to set up the registers as if a CPU reset had
307    occurred; hopefully real BIOSs don't assume much. */
308 static const unsigned char real_mode_switch [] =
309 {
310 	0x66, 0x0f, 0x20, 0xc0,			/*    movl  %cr0,%eax        */
311 	0x66, 0x83, 0xe0, 0x11,			/*    andl  $0x00000011,%eax */
312 	0x66, 0x0d, 0x00, 0x00, 0x00, 0x60,	/*    orl   $0x60000000,%eax */
313 	0x66, 0x0f, 0x22, 0xc0,			/*    movl  %eax,%cr0        */
314 	0x66, 0x0f, 0x22, 0xd8,			/*    movl  %eax,%cr3        */
315 	0x66, 0x0f, 0x20, 0xc3,			/*    movl  %cr0,%ebx        */
316 	0x66, 0x81, 0xe3, 0x00, 0x00, 0x00, 0x60,	/*    andl  $0x60000000,%ebx */
317 	0x74, 0x02,				/*    jz    f                */
318 	0x0f, 0x09,				/*    wbinvd                 */
319 	0x24, 0x10,				/* f: andb  $0x10,al         */
320 	0x66, 0x0f, 0x22, 0xc0			/*    movl  %eax,%cr0        */
321 };
322 static const unsigned char jump_to_bios [] =
323 {
324 	0xea, 0x00, 0x00, 0xff, 0xff		/*    ljmp  $0xffff,$0x0000  */
325 };
326 
327 /*
328  * Switch to real mode and then execute the code
329  * specified by the code and length parameters.
330  * We assume that length will aways be less that 100!
331  */
332 void machine_real_restart(const unsigned char *code, int length)
333 {
334 	local_irq_disable();
335 
336 	/* Write zero to CMOS register number 0x0f, which the BIOS POST
337 	   routine will recognize as telling it to do a proper reboot.  (Well
338 	   that's what this book in front of me says -- it may only apply to
339 	   the Phoenix BIOS though, it's not clear).  At the same time,
340 	   disable NMIs by setting the top bit in the CMOS address register,
341 	   as we're about to do peculiar things to the CPU.  I'm not sure if
342 	   `outb_p' is needed instead of just `outb'.  Use it to be on the
343 	   safe side.  (Yes, CMOS_WRITE does outb_p's. -  Paul G.)
344 	 */
345 	spin_lock(&rtc_lock);
346 	CMOS_WRITE(0x00, 0x8f);
347 	spin_unlock(&rtc_lock);
348 
349 	/* Remap the kernel at virtual address zero, as well as offset zero
350 	   from the kernel segment.  This assumes the kernel segment starts at
351 	   virtual address PAGE_OFFSET. */
352 	memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
353 		sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);
354 
355 	/*
356 	 * Use `swapper_pg_dir' as our page directory.
357 	 */
358 	load_cr3(swapper_pg_dir);
359 
360 	/* Write 0x1234 to absolute memory location 0x472.  The BIOS reads
361 	   this on booting to tell it to "Bypass memory test (also warm
362 	   boot)".  This seems like a fairly standard thing that gets set by
363 	   REBOOT.COM programs, and the previous reset routine did this
364 	   too. */
365 	*((unsigned short *)0x472) = reboot_mode;
366 
367 	/* For the switch to real mode, copy some code to low memory.  It has
368 	   to be in the first 64k because it is running in 16-bit mode, and it
369 	   has to have the same physical and virtual address, because it turns
370 	   off paging.  Copy it near the end of the first page, out of the way
371 	   of BIOS variables. */
372 	memcpy((void *)(0x1000 - sizeof(real_mode_switch) - 100),
373 		real_mode_switch, sizeof (real_mode_switch));
374 	memcpy((void *)(0x1000 - 100), code, length);
375 
376 	/* Set up the IDT for real mode. */
377 	load_idt(&real_mode_idt);
378 
379 	/* Set up a GDT from which we can load segment descriptors for real
380 	   mode.  The GDT is not used in real mode; it is just needed here to
381 	   prepare the descriptors. */
382 	load_gdt(&real_mode_gdt);
383 
384 	/* Load the data segment registers, and thus the descriptors ready for
385 	   real mode.  The base address of each segment is 0x100, 16 times the
386 	   selector value being loaded here.  This is so that the segment
387 	   registers don't have to be reloaded after switching to real mode:
388 	   the values are consistent for real mode operation already. */
389 	__asm__ __volatile__ ("movl $0x0010,%%eax\n"
390 				"\tmovl %%eax,%%ds\n"
391 				"\tmovl %%eax,%%es\n"
392 				"\tmovl %%eax,%%fs\n"
393 				"\tmovl %%eax,%%gs\n"
394 				"\tmovl %%eax,%%ss" : : : "eax");
395 
396 	/* Jump to the 16-bit code that we copied earlier.  It disables paging
397 	   and the cache, switches to real mode, and jumps to the BIOS reset
398 	   entry point. */
399 	__asm__ __volatile__ ("ljmp $0x0008,%0"
400 				:
401 				: "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
402 }
403 #ifdef CONFIG_APM_MODULE
404 EXPORT_SYMBOL(machine_real_restart);
405 #endif
406 
407 #endif /* CONFIG_X86_32 */
408 
409 /*
410  * Some Apple MacBook and MacBookPro's needs reboot=p to be able to reboot
411  */
412 static int __init set_pci_reboot(const struct dmi_system_id *d)
413 {
414 	if (reboot_type != BOOT_CF9) {
415 		reboot_type = BOOT_CF9;
416 		printk(KERN_INFO "%s series board detected. "
417 		       "Selecting PCI-method for reboots.\n", d->ident);
418 	}
419 	return 0;
420 }
421 
422 static struct dmi_system_id __initdata pci_reboot_dmi_table[] = {
423 	{	/* Handle problems with rebooting on Apple MacBook5 */
424 		.callback = set_pci_reboot,
425 		.ident = "Apple MacBook5",
426 		.matches = {
427 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
428 			DMI_MATCH(DMI_PRODUCT_NAME, "MacBook5"),
429 		},
430 	},
431 	{	/* Handle problems with rebooting on Apple MacBookPro5 */
432 		.callback = set_pci_reboot,
433 		.ident = "Apple MacBookPro5",
434 		.matches = {
435 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
436 			DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5"),
437 		},
438 	},
439 	{	/* Handle problems with rebooting on Apple Macmini3,1 */
440 		.callback = set_pci_reboot,
441 		.ident = "Apple Macmini3,1",
442 		.matches = {
443 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
444 			DMI_MATCH(DMI_PRODUCT_NAME, "Macmini3,1"),
445 		},
446 	},
447 	{ }
448 };
449 
450 static int __init pci_reboot_init(void)
451 {
452 	dmi_check_system(pci_reboot_dmi_table);
453 	return 0;
454 }
455 core_initcall(pci_reboot_init);
456 
457 static inline void kb_wait(void)
458 {
459 	int i;
460 
461 	for (i = 0; i < 0x10000; i++) {
462 		if ((inb(0x64) & 0x02) == 0)
463 			break;
464 		udelay(2);
465 	}
466 }
467 
468 static void vmxoff_nmi(int cpu, struct die_args *args)
469 {
470 	cpu_emergency_vmxoff();
471 }
472 
473 /* Use NMIs as IPIs to tell all CPUs to disable virtualization
474  */
475 static void emergency_vmx_disable_all(void)
476 {
477 	/* Just make sure we won't change CPUs while doing this */
478 	local_irq_disable();
479 
480 	/* We need to disable VMX on all CPUs before rebooting, otherwise
481 	 * we risk hanging up the machine, because the CPU ignore INIT
482 	 * signals when VMX is enabled.
483 	 *
484 	 * We can't take any locks and we may be on an inconsistent
485 	 * state, so we use NMIs as IPIs to tell the other CPUs to disable
486 	 * VMX and halt.
487 	 *
488 	 * For safety, we will avoid running the nmi_shootdown_cpus()
489 	 * stuff unnecessarily, but we don't have a way to check
490 	 * if other CPUs have VMX enabled. So we will call it only if the
491 	 * CPU we are running on has VMX enabled.
492 	 *
493 	 * We will miss cases where VMX is not enabled on all CPUs. This
494 	 * shouldn't do much harm because KVM always enable VMX on all
495 	 * CPUs anyway. But we can miss it on the small window where KVM
496 	 * is still enabling VMX.
497 	 */
498 	if (cpu_has_vmx() && cpu_vmx_enabled()) {
499 		/* Disable VMX on this CPU.
500 		 */
501 		cpu_vmxoff();
502 
503 		/* Halt and disable VMX on the other CPUs */
504 		nmi_shootdown_cpus(vmxoff_nmi);
505 
506 	}
507 }
508 
509 
510 void __attribute__((weak)) mach_reboot_fixups(void)
511 {
512 }
513 
514 static void native_machine_emergency_restart(void)
515 {
516 	int i;
517 
518 	if (reboot_emergency)
519 		emergency_vmx_disable_all();
520 
521 	tboot_shutdown(TB_SHUTDOWN_REBOOT);
522 
523 	/* Tell the BIOS if we want cold or warm reboot */
524 	*((unsigned short *)__va(0x472)) = reboot_mode;
525 
526 	for (;;) {
527 		/* Could also try the reset bit in the Hammer NB */
528 		switch (reboot_type) {
529 		case BOOT_KBD:
530 			mach_reboot_fixups(); /* for board specific fixups */
531 
532 			for (i = 0; i < 10; i++) {
533 				kb_wait();
534 				udelay(50);
535 				outb(0xfe, 0x64); /* pulse reset low */
536 				udelay(50);
537 			}
538 
539 		case BOOT_TRIPLE:
540 			load_idt(&no_idt);
541 			__asm__ __volatile__("int3");
542 
543 			reboot_type = BOOT_KBD;
544 			break;
545 
546 #ifdef CONFIG_X86_32
547 		case BOOT_BIOS:
548 			machine_real_restart(jump_to_bios, sizeof(jump_to_bios));
549 
550 			reboot_type = BOOT_KBD;
551 			break;
552 #endif
553 
554 		case BOOT_ACPI:
555 			acpi_reboot();
556 			reboot_type = BOOT_KBD;
557 			break;
558 
559 		case BOOT_EFI:
560 			if (efi_enabled)
561 				efi.reset_system(reboot_mode ?
562 						 EFI_RESET_WARM :
563 						 EFI_RESET_COLD,
564 						 EFI_SUCCESS, 0, NULL);
565 			reboot_type = BOOT_KBD;
566 			break;
567 
568 		case BOOT_CF9:
569 			port_cf9_safe = true;
570 			/* fall through */
571 
572 		case BOOT_CF9_COND:
573 			if (port_cf9_safe) {
574 				u8 cf9 = inb(0xcf9) & ~6;
575 				outb(cf9|2, 0xcf9); /* Request hard reset */
576 				udelay(50);
577 				outb(cf9|6, 0xcf9); /* Actually do the reset */
578 				udelay(50);
579 			}
580 			reboot_type = BOOT_KBD;
581 			break;
582 		}
583 	}
584 }
585 
586 void native_machine_shutdown(void)
587 {
588 	/* Stop the cpus and apics */
589 #ifdef CONFIG_SMP
590 
591 	/* The boot cpu is always logical cpu 0 */
592 	int reboot_cpu_id = 0;
593 
594 #ifdef CONFIG_X86_32
595 	/* See if there has been given a command line override */
596 	if ((reboot_cpu != -1) && (reboot_cpu < nr_cpu_ids) &&
597 		cpu_online(reboot_cpu))
598 		reboot_cpu_id = reboot_cpu;
599 #endif
600 
601 	/* Make certain the cpu I'm about to reboot on is online */
602 	if (!cpu_online(reboot_cpu_id))
603 		reboot_cpu_id = smp_processor_id();
604 
605 	/* Make certain I only run on the appropriate processor */
606 	set_cpus_allowed_ptr(current, cpumask_of(reboot_cpu_id));
607 
608 	/* O.K Now that I'm on the appropriate processor,
609 	 * stop all of the others.
610 	 */
611 	smp_send_stop();
612 #endif
613 
614 	lapic_shutdown();
615 
616 #ifdef CONFIG_X86_IO_APIC
617 	disable_IO_APIC();
618 #endif
619 
620 #ifdef CONFIG_HPET_TIMER
621 	hpet_disable();
622 #endif
623 
624 #ifdef CONFIG_X86_64
625 	pci_iommu_shutdown();
626 #endif
627 }
628 
629 static void __machine_emergency_restart(int emergency)
630 {
631 	reboot_emergency = emergency;
632 	machine_ops.emergency_restart();
633 }
634 
635 static void native_machine_restart(char *__unused)
636 {
637 	printk("machine restart\n");
638 
639 	if (!reboot_force)
640 		machine_shutdown();
641 	__machine_emergency_restart(0);
642 }
643 
644 static void native_machine_halt(void)
645 {
646 	/* stop other cpus and apics */
647 	machine_shutdown();
648 
649 	tboot_shutdown(TB_SHUTDOWN_HALT);
650 
651 	/* stop this cpu */
652 	stop_this_cpu(NULL);
653 }
654 
655 static void native_machine_power_off(void)
656 {
657 	if (pm_power_off) {
658 		if (!reboot_force)
659 			machine_shutdown();
660 		pm_power_off();
661 	}
662 	/* a fallback in case there is no PM info available */
663 	tboot_shutdown(TB_SHUTDOWN_HALT);
664 }
665 
666 struct machine_ops machine_ops = {
667 	.power_off = native_machine_power_off,
668 	.shutdown = native_machine_shutdown,
669 	.emergency_restart = native_machine_emergency_restart,
670 	.restart = native_machine_restart,
671 	.halt = native_machine_halt,
672 #ifdef CONFIG_KEXEC
673 	.crash_shutdown = native_machine_crash_shutdown,
674 #endif
675 };
676 
677 void machine_power_off(void)
678 {
679 	machine_ops.power_off();
680 }
681 
682 void machine_shutdown(void)
683 {
684 	machine_ops.shutdown();
685 }
686 
687 void machine_emergency_restart(void)
688 {
689 	__machine_emergency_restart(1);
690 }
691 
692 void machine_restart(char *cmd)
693 {
694 	machine_ops.restart(cmd);
695 }
696 
697 void machine_halt(void)
698 {
699 	machine_ops.halt();
700 }
701 
702 #ifdef CONFIG_KEXEC
703 void machine_crash_shutdown(struct pt_regs *regs)
704 {
705 	machine_ops.crash_shutdown(regs);
706 }
707 #endif
708 
709 
710 #if defined(CONFIG_SMP)
711 
712 /* This keeps a track of which one is crashing cpu. */
713 static int crashing_cpu;
714 static nmi_shootdown_cb shootdown_callback;
715 
716 static atomic_t waiting_for_crash_ipi;
717 
718 static int crash_nmi_callback(struct notifier_block *self,
719 			unsigned long val, void *data)
720 {
721 	int cpu;
722 
723 	if (val != DIE_NMI_IPI)
724 		return NOTIFY_OK;
725 
726 	cpu = raw_smp_processor_id();
727 
728 	/* Don't do anything if this handler is invoked on crashing cpu.
729 	 * Otherwise, system will completely hang. Crashing cpu can get
730 	 * an NMI if system was initially booted with nmi_watchdog parameter.
731 	 */
732 	if (cpu == crashing_cpu)
733 		return NOTIFY_STOP;
734 	local_irq_disable();
735 
736 	shootdown_callback(cpu, (struct die_args *)data);
737 
738 	atomic_dec(&waiting_for_crash_ipi);
739 	/* Assume hlt works */
740 	halt();
741 	for (;;)
742 		cpu_relax();
743 
744 	return 1;
745 }
746 
747 static void smp_send_nmi_allbutself(void)
748 {
749 	apic->send_IPI_allbutself(NMI_VECTOR);
750 }
751 
752 static struct notifier_block crash_nmi_nb = {
753 	.notifier_call = crash_nmi_callback,
754 };
755 
756 /* Halt all other CPUs, calling the specified function on each of them
757  *
758  * This function can be used to halt all other CPUs on crash
759  * or emergency reboot time. The function passed as parameter
760  * will be called inside a NMI handler on all CPUs.
761  */
762 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
763 {
764 	unsigned long msecs;
765 	local_irq_disable();
766 
767 	/* Make a note of crashing cpu. Will be used in NMI callback.*/
768 	crashing_cpu = safe_smp_processor_id();
769 
770 	shootdown_callback = callback;
771 
772 	atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
773 	/* Would it be better to replace the trap vector here? */
774 	if (register_die_notifier(&crash_nmi_nb))
775 		return;		/* return what? */
776 	/* Ensure the new callback function is set before sending
777 	 * out the NMI
778 	 */
779 	wmb();
780 
781 	smp_send_nmi_allbutself();
782 
783 	msecs = 1000; /* Wait at most a second for the other cpus to stop */
784 	while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
785 		mdelay(1);
786 		msecs--;
787 	}
788 
789 	/* Leave the nmi callback set */
790 }
791 #else /* !CONFIG_SMP */
792 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
793 {
794 	/* No other CPUs to shoot down */
795 }
796 #endif
797