xref: /linux/arch/x86/kernel/reboot.c (revision 1a5aeeecd550ee4344cfba1791f1134739b16dc6)
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 <acpi/reboot.h>
8 #include <asm/io.h>
9 #include <asm/apic.h>
10 #include <asm/desc.h>
11 #include <asm/hpet.h>
12 #include <asm/pgtable.h>
13 #include <asm/proto.h>
14 #include <asm/reboot_fixups.h>
15 #include <asm/reboot.h>
16 #include <asm/pci_x86.h>
17 #include <asm/virtext.h>
18 #include <asm/cpu.h>
19 
20 #ifdef CONFIG_X86_32
21 # include <linux/ctype.h>
22 # include <linux/mc146818rtc.h>
23 #else
24 # include <asm/iommu.h>
25 #endif
26 
27 /*
28  * Power off function, if any
29  */
30 void (*pm_power_off)(void);
31 EXPORT_SYMBOL(pm_power_off);
32 
33 static const struct desc_ptr no_idt = {};
34 static int reboot_mode;
35 enum reboot_type reboot_type = BOOT_KBD;
36 int reboot_force;
37 
38 #if defined(CONFIG_X86_32) && defined(CONFIG_SMP)
39 static int reboot_cpu = -1;
40 #endif
41 
42 /* This is set if we need to go through the 'emergency' path.
43  * When machine_emergency_restart() is called, we may be on
44  * an inconsistent state and won't be able to do a clean cleanup
45  */
46 static int reboot_emergency;
47 
48 /* This is set by the PCI code if either type 1 or type 2 PCI is detected */
49 bool port_cf9_safe = false;
50 
51 /* reboot=b[ios] | s[mp] | t[riple] | k[bd] | e[fi] [, [w]arm | [c]old] | p[ci]
52    warm   Don't set the cold reboot flag
53    cold   Set the cold reboot flag
54    bios   Reboot by jumping through the BIOS (only for X86_32)
55    smp    Reboot by executing reset on BSP or other CPU (only for X86_32)
56    triple Force a triple fault (init)
57    kbd    Use the keyboard controller. cold reset (default)
58    acpi   Use the RESET_REG in the FADT
59    efi    Use efi reset_system runtime service
60    pci    Use the so-called "PCI reset register", CF9
61    force  Avoid anything that could hang.
62  */
63 static int __init reboot_setup(char *str)
64 {
65 	for (;;) {
66 		switch (*str) {
67 		case 'w':
68 			reboot_mode = 0x1234;
69 			break;
70 
71 		case 'c':
72 			reboot_mode = 0;
73 			break;
74 
75 #ifdef CONFIG_X86_32
76 #ifdef CONFIG_SMP
77 		case 's':
78 			if (isdigit(*(str+1))) {
79 				reboot_cpu = (int) (*(str+1) - '0');
80 				if (isdigit(*(str+2)))
81 					reboot_cpu = reboot_cpu*10 + (int)(*(str+2) - '0');
82 			}
83 				/* we will leave sorting out the final value
84 				   when we are ready to reboot, since we might not
85 				   have set up boot_cpu_id or smp_num_cpu */
86 			break;
87 #endif /* CONFIG_SMP */
88 
89 		case 'b':
90 #endif
91 		case 'a':
92 		case 'k':
93 		case 't':
94 		case 'e':
95 		case 'p':
96 			reboot_type = *str;
97 			break;
98 
99 		case 'f':
100 			reboot_force = 1;
101 			break;
102 		}
103 
104 		str = strchr(str, ',');
105 		if (str)
106 			str++;
107 		else
108 			break;
109 	}
110 	return 1;
111 }
112 
113 __setup("reboot=", reboot_setup);
114 
115 
116 #ifdef CONFIG_X86_32
117 /*
118  * Reboot options and system auto-detection code provided by
119  * Dell Inc. so their systems "just work". :-)
120  */
121 
122 /*
123  * Some machines require the "reboot=b"  commandline option,
124  * this quirk makes that automatic.
125  */
126 static int __init set_bios_reboot(const struct dmi_system_id *d)
127 {
128 	if (reboot_type != BOOT_BIOS) {
129 		reboot_type = BOOT_BIOS;
130 		printk(KERN_INFO "%s series board detected. Selecting BIOS-method for reboots.\n", d->ident);
131 	}
132 	return 0;
133 }
134 
135 static struct dmi_system_id __initdata reboot_dmi_table[] = {
136 	{	/* Handle problems with rebooting on Dell E520's */
137 		.callback = set_bios_reboot,
138 		.ident = "Dell E520",
139 		.matches = {
140 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
141 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell DM061"),
142 		},
143 	},
144 	{	/* Handle problems with rebooting on Dell 1300's */
145 		.callback = set_bios_reboot,
146 		.ident = "Dell PowerEdge 1300",
147 		.matches = {
148 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
149 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 1300/"),
150 		},
151 	},
152 	{	/* Handle problems with rebooting on Dell 300's */
153 		.callback = set_bios_reboot,
154 		.ident = "Dell PowerEdge 300",
155 		.matches = {
156 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
157 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 300/"),
158 		},
159 	},
160 	{       /* Handle problems with rebooting on Dell Optiplex 745's SFF*/
161 		.callback = set_bios_reboot,
162 		.ident = "Dell OptiPlex 745",
163 		.matches = {
164 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
165 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
166 		},
167 	},
168 	{       /* Handle problems with rebooting on Dell Optiplex 745's DFF*/
169 		.callback = set_bios_reboot,
170 		.ident = "Dell OptiPlex 745",
171 		.matches = {
172 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
173 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
174 			DMI_MATCH(DMI_BOARD_NAME, "0MM599"),
175 		},
176 	},
177 	{       /* Handle problems with rebooting on Dell Optiplex 745 with 0KW626 */
178 		.callback = set_bios_reboot,
179 		.ident = "Dell OptiPlex 745",
180 		.matches = {
181 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
182 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
183 			DMI_MATCH(DMI_BOARD_NAME, "0KW626"),
184 		},
185 	},
186 	{   /* Handle problems with rebooting on Dell Optiplex 330 with 0KP561 */
187 		.callback = set_bios_reboot,
188 		.ident = "Dell OptiPlex 330",
189 		.matches = {
190 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
191 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 330"),
192 			DMI_MATCH(DMI_BOARD_NAME, "0KP561"),
193 		},
194 	},
195 	{   /* Handle problems with rebooting on Dell Optiplex 360 with 0T656F */
196 		.callback = set_bios_reboot,
197 		.ident = "Dell OptiPlex 360",
198 		.matches = {
199 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
200 			DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 360"),
201 			DMI_MATCH(DMI_BOARD_NAME, "0T656F"),
202 		},
203 	},
204 	{	/* Handle problems with rebooting on Dell 2400's */
205 		.callback = set_bios_reboot,
206 		.ident = "Dell PowerEdge 2400",
207 		.matches = {
208 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
209 			DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"),
210 		},
211 	},
212 	{	/* Handle problems with rebooting on Dell T5400's */
213 		.callback = set_bios_reboot,
214 		.ident = "Dell Precision T5400",
215 		.matches = {
216 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
217 			DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"),
218 		},
219 	},
220 	{	/* Handle problems with rebooting on HP laptops */
221 		.callback = set_bios_reboot,
222 		.ident = "HP Compaq Laptop",
223 		.matches = {
224 			DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
225 			DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq"),
226 		},
227 	},
228 	{	/* Handle problems with rebooting on Dell XPS710 */
229 		.callback = set_bios_reboot,
230 		.ident = "Dell XPS710",
231 		.matches = {
232 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
233 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
234 		},
235 	},
236 	{	/* Handle problems with rebooting on Dell DXP061 */
237 		.callback = set_bios_reboot,
238 		.ident = "Dell DXP061",
239 		.matches = {
240 			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
241 			DMI_MATCH(DMI_PRODUCT_NAME, "Dell DXP061"),
242 		},
243 	},
244 	{	/* Handle problems with rebooting on Sony VGN-Z540N */
245 		.callback = set_bios_reboot,
246 		.ident = "Sony VGN-Z540N",
247 		.matches = {
248 			DMI_MATCH(DMI_SYS_VENDOR, "Sony Corporation"),
249 			DMI_MATCH(DMI_PRODUCT_NAME, "VGN-Z540N"),
250 		},
251 	},
252 	{	/* Handle problems with rebooting on CompuLab SBC-FITPC2 */
253 		.callback = set_bios_reboot,
254 		.ident = "CompuLab SBC-FITPC2",
255 		.matches = {
256 			DMI_MATCH(DMI_SYS_VENDOR, "CompuLab"),
257 			DMI_MATCH(DMI_PRODUCT_NAME, "SBC-FITPC2"),
258 		},
259 	},
260 	{ }
261 };
262 
263 static int __init reboot_init(void)
264 {
265 	dmi_check_system(reboot_dmi_table);
266 	return 0;
267 }
268 core_initcall(reboot_init);
269 
270 /* The following code and data reboots the machine by switching to real
271    mode and jumping to the BIOS reset entry point, as if the CPU has
272    really been reset.  The previous version asked the keyboard
273    controller to pulse the CPU reset line, which is more thorough, but
274    doesn't work with at least one type of 486 motherboard.  It is easy
275    to stop this code working; hence the copious comments. */
276 static const unsigned long long
277 real_mode_gdt_entries [3] =
278 {
279 	0x0000000000000000ULL,	/* Null descriptor */
280 	0x00009b000000ffffULL,	/* 16-bit real-mode 64k code at 0x00000000 */
281 	0x000093000100ffffULL	/* 16-bit real-mode 64k data at 0x00000100 */
282 };
283 
284 static const struct desc_ptr
285 real_mode_gdt = { sizeof (real_mode_gdt_entries) - 1, (long)real_mode_gdt_entries },
286 real_mode_idt = { 0x3ff, 0 };
287 
288 /* This is 16-bit protected mode code to disable paging and the cache,
289    switch to real mode and jump to the BIOS reset code.
290 
291    The instruction that switches to real mode by writing to CR0 must be
292    followed immediately by a far jump instruction, which set CS to a
293    valid value for real mode, and flushes the prefetch queue to avoid
294    running instructions that have already been decoded in protected
295    mode.
296 
297    Clears all the flags except ET, especially PG (paging), PE
298    (protected-mode enable) and TS (task switch for coprocessor state
299    save).  Flushes the TLB after paging has been disabled.  Sets CD and
300    NW, to disable the cache on a 486, and invalidates the cache.  This
301    is more like the state of a 486 after reset.  I don't know if
302    something else should be done for other chips.
303 
304    More could be done here to set up the registers as if a CPU reset had
305    occurred; hopefully real BIOSs don't assume much. */
306 static const unsigned char real_mode_switch [] =
307 {
308 	0x66, 0x0f, 0x20, 0xc0,			/*    movl  %cr0,%eax        */
309 	0x66, 0x83, 0xe0, 0x11,			/*    andl  $0x00000011,%eax */
310 	0x66, 0x0d, 0x00, 0x00, 0x00, 0x60,	/*    orl   $0x60000000,%eax */
311 	0x66, 0x0f, 0x22, 0xc0,			/*    movl  %eax,%cr0        */
312 	0x66, 0x0f, 0x22, 0xd8,			/*    movl  %eax,%cr3        */
313 	0x66, 0x0f, 0x20, 0xc3,			/*    movl  %cr0,%ebx        */
314 	0x66, 0x81, 0xe3, 0x00, 0x00, 0x00, 0x60,	/*    andl  $0x60000000,%ebx */
315 	0x74, 0x02,				/*    jz    f                */
316 	0x0f, 0x09,				/*    wbinvd                 */
317 	0x24, 0x10,				/* f: andb  $0x10,al         */
318 	0x66, 0x0f, 0x22, 0xc0			/*    movl  %eax,%cr0        */
319 };
320 static const unsigned char jump_to_bios [] =
321 {
322 	0xea, 0x00, 0x00, 0xff, 0xff		/*    ljmp  $0xffff,$0x0000  */
323 };
324 
325 /*
326  * Switch to real mode and then execute the code
327  * specified by the code and length parameters.
328  * We assume that length will aways be less that 100!
329  */
330 void machine_real_restart(const unsigned char *code, int length)
331 {
332 	local_irq_disable();
333 
334 	/* Write zero to CMOS register number 0x0f, which the BIOS POST
335 	   routine will recognize as telling it to do a proper reboot.  (Well
336 	   that's what this book in front of me says -- it may only apply to
337 	   the Phoenix BIOS though, it's not clear).  At the same time,
338 	   disable NMIs by setting the top bit in the CMOS address register,
339 	   as we're about to do peculiar things to the CPU.  I'm not sure if
340 	   `outb_p' is needed instead of just `outb'.  Use it to be on the
341 	   safe side.  (Yes, CMOS_WRITE does outb_p's. -  Paul G.)
342 	 */
343 	spin_lock(&rtc_lock);
344 	CMOS_WRITE(0x00, 0x8f);
345 	spin_unlock(&rtc_lock);
346 
347 	/* Remap the kernel at virtual address zero, as well as offset zero
348 	   from the kernel segment.  This assumes the kernel segment starts at
349 	   virtual address PAGE_OFFSET. */
350 	memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
351 		sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);
352 
353 	/*
354 	 * Use `swapper_pg_dir' as our page directory.
355 	 */
356 	load_cr3(swapper_pg_dir);
357 
358 	/* Write 0x1234 to absolute memory location 0x472.  The BIOS reads
359 	   this on booting to tell it to "Bypass memory test (also warm
360 	   boot)".  This seems like a fairly standard thing that gets set by
361 	   REBOOT.COM programs, and the previous reset routine did this
362 	   too. */
363 	*((unsigned short *)0x472) = reboot_mode;
364 
365 	/* For the switch to real mode, copy some code to low memory.  It has
366 	   to be in the first 64k because it is running in 16-bit mode, and it
367 	   has to have the same physical and virtual address, because it turns
368 	   off paging.  Copy it near the end of the first page, out of the way
369 	   of BIOS variables. */
370 	memcpy((void *)(0x1000 - sizeof(real_mode_switch) - 100),
371 		real_mode_switch, sizeof (real_mode_switch));
372 	memcpy((void *)(0x1000 - 100), code, length);
373 
374 	/* Set up the IDT for real mode. */
375 	load_idt(&real_mode_idt);
376 
377 	/* Set up a GDT from which we can load segment descriptors for real
378 	   mode.  The GDT is not used in real mode; it is just needed here to
379 	   prepare the descriptors. */
380 	load_gdt(&real_mode_gdt);
381 
382 	/* Load the data segment registers, and thus the descriptors ready for
383 	   real mode.  The base address of each segment is 0x100, 16 times the
384 	   selector value being loaded here.  This is so that the segment
385 	   registers don't have to be reloaded after switching to real mode:
386 	   the values are consistent for real mode operation already. */
387 	__asm__ __volatile__ ("movl $0x0010,%%eax\n"
388 				"\tmovl %%eax,%%ds\n"
389 				"\tmovl %%eax,%%es\n"
390 				"\tmovl %%eax,%%fs\n"
391 				"\tmovl %%eax,%%gs\n"
392 				"\tmovl %%eax,%%ss" : : : "eax");
393 
394 	/* Jump to the 16-bit code that we copied earlier.  It disables paging
395 	   and the cache, switches to real mode, and jumps to the BIOS reset
396 	   entry point. */
397 	__asm__ __volatile__ ("ljmp $0x0008,%0"
398 				:
399 				: "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
400 }
401 #ifdef CONFIG_APM_MODULE
402 EXPORT_SYMBOL(machine_real_restart);
403 #endif
404 
405 #endif /* CONFIG_X86_32 */
406 
407 /*
408  * Apple MacBook5,2 (2009 MacBook) needs reboot=p
409  */
410 static int __init set_pci_reboot(const struct dmi_system_id *d)
411 {
412 	if (reboot_type != BOOT_CF9) {
413 		reboot_type = BOOT_CF9;
414 		printk(KERN_INFO "%s series board detected. "
415 		       "Selecting PCI-method for reboots.\n", d->ident);
416 	}
417 	return 0;
418 }
419 
420 static struct dmi_system_id __initdata pci_reboot_dmi_table[] = {
421 	{	/* Handle problems with rebooting on Apple MacBook5,2 */
422 		.callback = set_pci_reboot,
423 		.ident = "Apple MacBook",
424 		.matches = {
425 			DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
426 			DMI_MATCH(DMI_PRODUCT_NAME, "MacBook5,2"),
427 		},
428 	},
429 	{ }
430 };
431 
432 static int __init pci_reboot_init(void)
433 {
434 	dmi_check_system(pci_reboot_dmi_table);
435 	return 0;
436 }
437 core_initcall(pci_reboot_init);
438 
439 static inline void kb_wait(void)
440 {
441 	int i;
442 
443 	for (i = 0; i < 0x10000; i++) {
444 		if ((inb(0x64) & 0x02) == 0)
445 			break;
446 		udelay(2);
447 	}
448 }
449 
450 static void vmxoff_nmi(int cpu, struct die_args *args)
451 {
452 	cpu_emergency_vmxoff();
453 }
454 
455 /* Use NMIs as IPIs to tell all CPUs to disable virtualization
456  */
457 static void emergency_vmx_disable_all(void)
458 {
459 	/* Just make sure we won't change CPUs while doing this */
460 	local_irq_disable();
461 
462 	/* We need to disable VMX on all CPUs before rebooting, otherwise
463 	 * we risk hanging up the machine, because the CPU ignore INIT
464 	 * signals when VMX is enabled.
465 	 *
466 	 * We can't take any locks and we may be on an inconsistent
467 	 * state, so we use NMIs as IPIs to tell the other CPUs to disable
468 	 * VMX and halt.
469 	 *
470 	 * For safety, we will avoid running the nmi_shootdown_cpus()
471 	 * stuff unnecessarily, but we don't have a way to check
472 	 * if other CPUs have VMX enabled. So we will call it only if the
473 	 * CPU we are running on has VMX enabled.
474 	 *
475 	 * We will miss cases where VMX is not enabled on all CPUs. This
476 	 * shouldn't do much harm because KVM always enable VMX on all
477 	 * CPUs anyway. But we can miss it on the small window where KVM
478 	 * is still enabling VMX.
479 	 */
480 	if (cpu_has_vmx() && cpu_vmx_enabled()) {
481 		/* Disable VMX on this CPU.
482 		 */
483 		cpu_vmxoff();
484 
485 		/* Halt and disable VMX on the other CPUs */
486 		nmi_shootdown_cpus(vmxoff_nmi);
487 
488 	}
489 }
490 
491 
492 void __attribute__((weak)) mach_reboot_fixups(void)
493 {
494 }
495 
496 static void native_machine_emergency_restart(void)
497 {
498 	int i;
499 
500 	if (reboot_emergency)
501 		emergency_vmx_disable_all();
502 
503 	/* Tell the BIOS if we want cold or warm reboot */
504 	*((unsigned short *)__va(0x472)) = reboot_mode;
505 
506 	for (;;) {
507 		/* Could also try the reset bit in the Hammer NB */
508 		switch (reboot_type) {
509 		case BOOT_KBD:
510 			mach_reboot_fixups(); /* for board specific fixups */
511 
512 			for (i = 0; i < 10; i++) {
513 				kb_wait();
514 				udelay(50);
515 				outb(0xfe, 0x64); /* pulse reset low */
516 				udelay(50);
517 			}
518 
519 		case BOOT_TRIPLE:
520 			load_idt(&no_idt);
521 			__asm__ __volatile__("int3");
522 
523 			reboot_type = BOOT_KBD;
524 			break;
525 
526 #ifdef CONFIG_X86_32
527 		case BOOT_BIOS:
528 			machine_real_restart(jump_to_bios, sizeof(jump_to_bios));
529 
530 			reboot_type = BOOT_KBD;
531 			break;
532 #endif
533 
534 		case BOOT_ACPI:
535 			acpi_reboot();
536 			reboot_type = BOOT_KBD;
537 			break;
538 
539 		case BOOT_EFI:
540 			if (efi_enabled)
541 				efi.reset_system(reboot_mode ?
542 						 EFI_RESET_WARM :
543 						 EFI_RESET_COLD,
544 						 EFI_SUCCESS, 0, NULL);
545 			reboot_type = BOOT_KBD;
546 			break;
547 
548 		case BOOT_CF9:
549 			port_cf9_safe = true;
550 			/* fall through */
551 
552 		case BOOT_CF9_COND:
553 			if (port_cf9_safe) {
554 				u8 cf9 = inb(0xcf9) & ~6;
555 				outb(cf9|2, 0xcf9); /* Request hard reset */
556 				udelay(50);
557 				outb(cf9|6, 0xcf9); /* Actually do the reset */
558 				udelay(50);
559 			}
560 			reboot_type = BOOT_KBD;
561 			break;
562 		}
563 	}
564 }
565 
566 void native_machine_shutdown(void)
567 {
568 	/* Stop the cpus and apics */
569 #ifdef CONFIG_SMP
570 
571 	/* The boot cpu is always logical cpu 0 */
572 	int reboot_cpu_id = 0;
573 
574 #ifdef CONFIG_X86_32
575 	/* See if there has been given a command line override */
576 	if ((reboot_cpu != -1) && (reboot_cpu < nr_cpu_ids) &&
577 		cpu_online(reboot_cpu))
578 		reboot_cpu_id = reboot_cpu;
579 #endif
580 
581 	/* Make certain the cpu I'm about to reboot on is online */
582 	if (!cpu_online(reboot_cpu_id))
583 		reboot_cpu_id = smp_processor_id();
584 
585 	/* Make certain I only run on the appropriate processor */
586 	set_cpus_allowed_ptr(current, cpumask_of(reboot_cpu_id));
587 
588 	/* O.K Now that I'm on the appropriate processor,
589 	 * stop all of the others.
590 	 */
591 	smp_send_stop();
592 #endif
593 
594 	lapic_shutdown();
595 
596 #ifdef CONFIG_X86_IO_APIC
597 	disable_IO_APIC();
598 #endif
599 
600 #ifdef CONFIG_HPET_TIMER
601 	hpet_disable();
602 #endif
603 
604 #ifdef CONFIG_X86_64
605 	pci_iommu_shutdown();
606 #endif
607 }
608 
609 static void __machine_emergency_restart(int emergency)
610 {
611 	reboot_emergency = emergency;
612 	machine_ops.emergency_restart();
613 }
614 
615 static void native_machine_restart(char *__unused)
616 {
617 	printk("machine restart\n");
618 
619 	if (!reboot_force)
620 		machine_shutdown();
621 	__machine_emergency_restart(0);
622 }
623 
624 static void native_machine_halt(void)
625 {
626 	/* stop other cpus and apics */
627 	machine_shutdown();
628 
629 	/* stop this cpu */
630 	stop_this_cpu(NULL);
631 }
632 
633 static void native_machine_power_off(void)
634 {
635 	if (pm_power_off) {
636 		if (!reboot_force)
637 			machine_shutdown();
638 		pm_power_off();
639 	}
640 }
641 
642 struct machine_ops machine_ops = {
643 	.power_off = native_machine_power_off,
644 	.shutdown = native_machine_shutdown,
645 	.emergency_restart = native_machine_emergency_restart,
646 	.restart = native_machine_restart,
647 	.halt = native_machine_halt,
648 #ifdef CONFIG_KEXEC
649 	.crash_shutdown = native_machine_crash_shutdown,
650 #endif
651 };
652 
653 void machine_power_off(void)
654 {
655 	machine_ops.power_off();
656 }
657 
658 void machine_shutdown(void)
659 {
660 	machine_ops.shutdown();
661 }
662 
663 void machine_emergency_restart(void)
664 {
665 	__machine_emergency_restart(1);
666 }
667 
668 void machine_restart(char *cmd)
669 {
670 	machine_ops.restart(cmd);
671 }
672 
673 void machine_halt(void)
674 {
675 	machine_ops.halt();
676 }
677 
678 #ifdef CONFIG_KEXEC
679 void machine_crash_shutdown(struct pt_regs *regs)
680 {
681 	machine_ops.crash_shutdown(regs);
682 }
683 #endif
684 
685 
686 #if defined(CONFIG_SMP)
687 
688 /* This keeps a track of which one is crashing cpu. */
689 static int crashing_cpu;
690 static nmi_shootdown_cb shootdown_callback;
691 
692 static atomic_t waiting_for_crash_ipi;
693 
694 static int crash_nmi_callback(struct notifier_block *self,
695 			unsigned long val, void *data)
696 {
697 	int cpu;
698 
699 	if (val != DIE_NMI_IPI)
700 		return NOTIFY_OK;
701 
702 	cpu = raw_smp_processor_id();
703 
704 	/* Don't do anything if this handler is invoked on crashing cpu.
705 	 * Otherwise, system will completely hang. Crashing cpu can get
706 	 * an NMI if system was initially booted with nmi_watchdog parameter.
707 	 */
708 	if (cpu == crashing_cpu)
709 		return NOTIFY_STOP;
710 	local_irq_disable();
711 
712 	shootdown_callback(cpu, (struct die_args *)data);
713 
714 	atomic_dec(&waiting_for_crash_ipi);
715 	/* Assume hlt works */
716 	halt();
717 	for (;;)
718 		cpu_relax();
719 
720 	return 1;
721 }
722 
723 static void smp_send_nmi_allbutself(void)
724 {
725 	apic->send_IPI_allbutself(NMI_VECTOR);
726 }
727 
728 static struct notifier_block crash_nmi_nb = {
729 	.notifier_call = crash_nmi_callback,
730 };
731 
732 /* Halt all other CPUs, calling the specified function on each of them
733  *
734  * This function can be used to halt all other CPUs on crash
735  * or emergency reboot time. The function passed as parameter
736  * will be called inside a NMI handler on all CPUs.
737  */
738 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
739 {
740 	unsigned long msecs;
741 	local_irq_disable();
742 
743 	/* Make a note of crashing cpu. Will be used in NMI callback.*/
744 	crashing_cpu = safe_smp_processor_id();
745 
746 	shootdown_callback = callback;
747 
748 	atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
749 	/* Would it be better to replace the trap vector here? */
750 	if (register_die_notifier(&crash_nmi_nb))
751 		return;		/* return what? */
752 	/* Ensure the new callback function is set before sending
753 	 * out the NMI
754 	 */
755 	wmb();
756 
757 	smp_send_nmi_allbutself();
758 
759 	msecs = 1000; /* Wait at most a second for the other cpus to stop */
760 	while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
761 		mdelay(1);
762 		msecs--;
763 	}
764 
765 	/* Leave the nmi callback set */
766 }
767 #else /* !CONFIG_SMP */
768 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
769 {
770 	/* No other CPUs to shoot down */
771 }
772 #endif
773