1 #include <linux/module.h>
2 #include <linux/reboot.h>
3 #include <linux/init.h>
7 #include <linux/tboot.h>
8 #include <acpi/reboot.h>
13 #include <asm/pgtable.h>
14 #include <asm/proto.h>
15 #include <asm/reboot_fixups.h>
16 #include <asm/reboot.h>
17 #include <asm/pci_x86.h>
18 #include <asm/virtext.h>
22 # include <linux/ctype.h>
23 # include <linux/mc146818rtc.h>
25 # include <asm/iommu.h>
29 * Power off function, if any
31 void (*pm_power_off)(void);
32 EXPORT_SYMBOL(pm_power_off);
34 static const struct desc_ptr no_idt = {};
35 static int reboot_mode;
36 enum reboot_type reboot_type = BOOT_KBD;
39 #if defined(CONFIG_X86_32) && defined(CONFIG_SMP)
40 static int reboot_cpu = -1;
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
47 static int reboot_emergency;
49 /* This is set by the PCI code if either type 1 or type 2 PCI is detected */
50 bool port_cf9_safe = false;
52 /* reboot=b[ios] | s[mp] | t[riple] | k[bd] | e[fi] [, [w]arm | [c]old] | p[ci]
53 warm Don't set the cold reboot flag
54 cold Set the cold reboot flag
55 bios Reboot by jumping through the BIOS (only for X86_32)
56 smp Reboot by executing reset on BSP or other CPU (only for X86_32)
57 triple Force a triple fault (init)
58 kbd Use the keyboard controller. cold reset (default)
59 acpi Use the RESET_REG in the FADT
60 efi Use efi reset_system runtime service
61 pci Use the so-called "PCI reset register", CF9
62 force Avoid anything that could hang.
64 static int __init reboot_setup(char *str)
79 if (isdigit(*(str+1))) {
80 reboot_cpu = (int) (*(str+1) - '0');
81 if (isdigit(*(str+2)))
82 reboot_cpu = reboot_cpu*10 + (int)(*(str+2) - '0');
84 /* we will leave sorting out the final value
85 when we are ready to reboot, since we might not
86 have set up boot_cpu_id or smp_num_cpu */
88 #endif /* CONFIG_SMP */
105 str = strchr(str, ',');
114 __setup("reboot=", reboot_setup);
119 * Reboot options and system auto-detection code provided by
120 * Dell Inc. so their systems "just work". :-)
124 * Some machines require the "reboot=b" commandline option,
125 * this quirk makes that automatic.
127 static int __init set_bios_reboot(const struct dmi_system_id *d)
129 if (reboot_type != BOOT_BIOS) {
130 reboot_type = BOOT_BIOS;
131 printk(KERN_INFO "%s series board detected. Selecting BIOS-method for reboots.\n", d->ident);
136 static struct dmi_system_id __initdata reboot_dmi_table[] = {
137 { /* Handle problems with rebooting on Dell E520's */
138 .callback = set_bios_reboot,
139 .ident = "Dell E520",
141 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
142 DMI_MATCH(DMI_PRODUCT_NAME, "Dell DM061"),
145 { /* Handle problems with rebooting on Dell 1300's */
146 .callback = set_bios_reboot,
147 .ident = "Dell PowerEdge 1300",
149 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
150 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 1300/"),
153 { /* Handle problems with rebooting on Dell 300's */
154 .callback = set_bios_reboot,
155 .ident = "Dell PowerEdge 300",
157 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
158 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 300/"),
161 { /* Handle problems with rebooting on Dell Optiplex 745's SFF*/
162 .callback = set_bios_reboot,
163 .ident = "Dell OptiPlex 745",
165 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
166 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
169 { /* Handle problems with rebooting on Dell Optiplex 745's DFF*/
170 .callback = set_bios_reboot,
171 .ident = "Dell OptiPlex 745",
173 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
174 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
175 DMI_MATCH(DMI_BOARD_NAME, "0MM599"),
178 { /* Handle problems with rebooting on Dell Optiplex 745 with 0KW626 */
179 .callback = set_bios_reboot,
180 .ident = "Dell OptiPlex 745",
182 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
183 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
184 DMI_MATCH(DMI_BOARD_NAME, "0KW626"),
187 { /* Handle problems with rebooting on Dell Optiplex 330 with 0KP561 */
188 .callback = set_bios_reboot,
189 .ident = "Dell OptiPlex 330",
191 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
192 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 330"),
193 DMI_MATCH(DMI_BOARD_NAME, "0KP561"),
196 { /* Handle problems with rebooting on Dell Optiplex 360 with 0T656F */
197 .callback = set_bios_reboot,
198 .ident = "Dell OptiPlex 360",
200 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
201 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 360"),
202 DMI_MATCH(DMI_BOARD_NAME, "0T656F"),
205 { /* Handle problems with rebooting on Dell 2400's */
206 .callback = set_bios_reboot,
207 .ident = "Dell PowerEdge 2400",
209 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
210 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"),
213 { /* Handle problems with rebooting on Dell T5400's */
214 .callback = set_bios_reboot,
215 .ident = "Dell Precision T5400",
217 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
218 DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"),
221 { /* Handle problems with rebooting on HP laptops */
222 .callback = set_bios_reboot,
223 .ident = "HP Compaq Laptop",
225 DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
226 DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq"),
229 { /* Handle problems with rebooting on Dell XPS710 */
230 .callback = set_bios_reboot,
231 .ident = "Dell XPS710",
233 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
234 DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
237 { /* Handle problems with rebooting on Dell DXP061 */
238 .callback = set_bios_reboot,
239 .ident = "Dell DXP061",
241 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
242 DMI_MATCH(DMI_PRODUCT_NAME, "Dell DXP061"),
245 { /* Handle problems with rebooting on Sony VGN-Z540N */
246 .callback = set_bios_reboot,
247 .ident = "Sony VGN-Z540N",
249 DMI_MATCH(DMI_SYS_VENDOR, "Sony Corporation"),
250 DMI_MATCH(DMI_PRODUCT_NAME, "VGN-Z540N"),
253 { /* Handle problems with rebooting on CompuLab SBC-FITPC2 */
254 .callback = set_bios_reboot,
255 .ident = "CompuLab SBC-FITPC2",
257 DMI_MATCH(DMI_SYS_VENDOR, "CompuLab"),
258 DMI_MATCH(DMI_PRODUCT_NAME, "SBC-FITPC2"),
264 static int __init reboot_init(void)
266 dmi_check_system(reboot_dmi_table);
269 core_initcall(reboot_init);
271 /* The following code and data reboots the machine by switching to real
272 mode and jumping to the BIOS reset entry point, as if the CPU has
273 really been reset. The previous version asked the keyboard
274 controller to pulse the CPU reset line, which is more thorough, but
275 doesn't work with at least one type of 486 motherboard. It is easy
276 to stop this code working; hence the copious comments. */
277 static const unsigned long long
278 real_mode_gdt_entries [3] =
280 0x0000000000000000ULL, /* Null descriptor */
281 0x00009b000000ffffULL, /* 16-bit real-mode 64k code at 0x00000000 */
282 0x000093000100ffffULL /* 16-bit real-mode 64k data at 0x00000100 */
285 static const struct desc_ptr
286 real_mode_gdt = { sizeof (real_mode_gdt_entries) - 1, (long)real_mode_gdt_entries },
287 real_mode_idt = { 0x3ff, 0 };
289 /* This is 16-bit protected mode code to disable paging and the cache,
290 switch to real mode and jump to the BIOS reset code.
292 The instruction that switches to real mode by writing to CR0 must be
293 followed immediately by a far jump instruction, which set CS to a
294 valid value for real mode, and flushes the prefetch queue to avoid
295 running instructions that have already been decoded in protected
298 Clears all the flags except ET, especially PG (paging), PE
299 (protected-mode enable) and TS (task switch for coprocessor state
300 save). Flushes the TLB after paging has been disabled. Sets CD and
301 NW, to disable the cache on a 486, and invalidates the cache. This
302 is more like the state of a 486 after reset. I don't know if
303 something else should be done for other chips.
305 More could be done here to set up the registers as if a CPU reset had
306 occurred; hopefully real BIOSs don't assume much. */
307 static const unsigned char real_mode_switch [] =
309 0x66, 0x0f, 0x20, 0xc0, /* movl %cr0,%eax */
310 0x66, 0x83, 0xe0, 0x11, /* andl $0x00000011,%eax */
311 0x66, 0x0d, 0x00, 0x00, 0x00, 0x60, /* orl $0x60000000,%eax */
312 0x66, 0x0f, 0x22, 0xc0, /* movl %eax,%cr0 */
313 0x66, 0x0f, 0x22, 0xd8, /* movl %eax,%cr3 */
314 0x66, 0x0f, 0x20, 0xc3, /* movl %cr0,%ebx */
315 0x66, 0x81, 0xe3, 0x00, 0x00, 0x00, 0x60, /* andl $0x60000000,%ebx */
316 0x74, 0x02, /* jz f */
317 0x0f, 0x09, /* wbinvd */
318 0x24, 0x10, /* f: andb $0x10,al */
319 0x66, 0x0f, 0x22, 0xc0 /* movl %eax,%cr0 */
321 static const unsigned char jump_to_bios [] =
323 0xea, 0x00, 0x00, 0xff, 0xff /* ljmp $0xffff,$0x0000 */
327 * Switch to real mode and then execute the code
328 * specified by the code and length parameters.
329 * We assume that length will aways be less that 100!
331 void machine_real_restart(const unsigned char *code, int length)
335 /* Write zero to CMOS register number 0x0f, which the BIOS POST
336 routine will recognize as telling it to do a proper reboot. (Well
337 that's what this book in front of me says -- it may only apply to
338 the Phoenix BIOS though, it's not clear). At the same time,
339 disable NMIs by setting the top bit in the CMOS address register,
340 as we're about to do peculiar things to the CPU. I'm not sure if
341 `outb_p' is needed instead of just `outb'. Use it to be on the
342 safe side. (Yes, CMOS_WRITE does outb_p's. - Paul G.)
344 spin_lock(&rtc_lock);
345 CMOS_WRITE(0x00, 0x8f);
346 spin_unlock(&rtc_lock);
348 /* Remap the kernel at virtual address zero, as well as offset zero
349 from the kernel segment. This assumes the kernel segment starts at
350 virtual address PAGE_OFFSET. */
351 memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
352 sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);
355 * Use `swapper_pg_dir' as our page directory.
357 load_cr3(swapper_pg_dir);
359 /* Write 0x1234 to absolute memory location 0x472. The BIOS reads
360 this on booting to tell it to "Bypass memory test (also warm
361 boot)". This seems like a fairly standard thing that gets set by
362 REBOOT.COM programs, and the previous reset routine did this
364 *((unsigned short *)0x472) = reboot_mode;
366 /* For the switch to real mode, copy some code to low memory. It has
367 to be in the first 64k because it is running in 16-bit mode, and it
368 has to have the same physical and virtual address, because it turns
369 off paging. Copy it near the end of the first page, out of the way
370 of BIOS variables. */
371 memcpy((void *)(0x1000 - sizeof(real_mode_switch) - 100),
372 real_mode_switch, sizeof (real_mode_switch));
373 memcpy((void *)(0x1000 - 100), code, length);
375 /* Set up the IDT for real mode. */
376 load_idt(&real_mode_idt);
378 /* Set up a GDT from which we can load segment descriptors for real
379 mode. The GDT is not used in real mode; it is just needed here to
380 prepare the descriptors. */
381 load_gdt(&real_mode_gdt);
383 /* Load the data segment registers, and thus the descriptors ready for
384 real mode. The base address of each segment is 0x100, 16 times the
385 selector value being loaded here. This is so that the segment
386 registers don't have to be reloaded after switching to real mode:
387 the values are consistent for real mode operation already. */
388 __asm__ __volatile__ ("movl $0x0010,%%eax\n"
389 "\tmovl %%eax,%%ds\n"
390 "\tmovl %%eax,%%es\n"
391 "\tmovl %%eax,%%fs\n"
392 "\tmovl %%eax,%%gs\n"
393 "\tmovl %%eax,%%ss" : : : "eax");
395 /* Jump to the 16-bit code that we copied earlier. It disables paging
396 and the cache, switches to real mode, and jumps to the BIOS reset
398 __asm__ __volatile__ ("ljmp $0x0008,%0"
400 : "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
402 #ifdef CONFIG_APM_MODULE
403 EXPORT_SYMBOL(machine_real_restart);
406 #endif /* CONFIG_X86_32 */
409 * Some Apple MacBook and MacBookPro's needs reboot=p to be able to reboot
411 static int __init set_pci_reboot(const struct dmi_system_id *d)
413 if (reboot_type != BOOT_CF9) {
414 reboot_type = BOOT_CF9;
415 printk(KERN_INFO "%s series board detected. "
416 "Selecting PCI-method for reboots.\n", d->ident);
421 static struct dmi_system_id __initdata pci_reboot_dmi_table[] = {
422 { /* Handle problems with rebooting on Apple MacBook5 */
423 .callback = set_pci_reboot,
424 .ident = "Apple MacBook5",
426 DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
427 DMI_MATCH(DMI_PRODUCT_NAME, "MacBook5"),
430 { /* Handle problems with rebooting on Apple MacBookPro5 */
431 .callback = set_pci_reboot,
432 .ident = "Apple MacBookPro5",
434 DMI_MATCH(DMI_SYS_VENDOR, "Apple Inc."),
435 DMI_MATCH(DMI_PRODUCT_NAME, "MacBookPro5"),
441 static int __init pci_reboot_init(void)
443 dmi_check_system(pci_reboot_dmi_table);
446 core_initcall(pci_reboot_init);
448 static inline void kb_wait(void)
452 for (i = 0; i < 0x10000; i++) {
453 if ((inb(0x64) & 0x02) == 0)
459 static void vmxoff_nmi(int cpu, struct die_args *args)
461 cpu_emergency_vmxoff();
464 /* Use NMIs as IPIs to tell all CPUs to disable virtualization
466 static void emergency_vmx_disable_all(void)
468 /* Just make sure we won't change CPUs while doing this */
471 /* We need to disable VMX on all CPUs before rebooting, otherwise
472 * we risk hanging up the machine, because the CPU ignore INIT
473 * signals when VMX is enabled.
475 * We can't take any locks and we may be on an inconsistent
476 * state, so we use NMIs as IPIs to tell the other CPUs to disable
479 * For safety, we will avoid running the nmi_shootdown_cpus()
480 * stuff unnecessarily, but we don't have a way to check
481 * if other CPUs have VMX enabled. So we will call it only if the
482 * CPU we are running on has VMX enabled.
484 * We will miss cases where VMX is not enabled on all CPUs. This
485 * shouldn't do much harm because KVM always enable VMX on all
486 * CPUs anyway. But we can miss it on the small window where KVM
487 * is still enabling VMX.
489 if (cpu_has_vmx() && cpu_vmx_enabled()) {
490 /* Disable VMX on this CPU.
494 /* Halt and disable VMX on the other CPUs */
495 nmi_shootdown_cpus(vmxoff_nmi);
501 void __attribute__((weak)) mach_reboot_fixups(void)
505 static void native_machine_emergency_restart(void)
509 if (reboot_emergency)
510 emergency_vmx_disable_all();
512 tboot_shutdown(TB_SHUTDOWN_REBOOT);
514 /* Tell the BIOS if we want cold or warm reboot */
515 *((unsigned short *)__va(0x472)) = reboot_mode;
518 /* Could also try the reset bit in the Hammer NB */
519 switch (reboot_type) {
521 mach_reboot_fixups(); /* for board specific fixups */
523 for (i = 0; i < 10; i++) {
526 outb(0xfe, 0x64); /* pulse reset low */
532 __asm__ __volatile__("int3");
534 reboot_type = BOOT_KBD;
539 machine_real_restart(jump_to_bios, sizeof(jump_to_bios));
541 reboot_type = BOOT_KBD;
547 reboot_type = BOOT_KBD;
552 efi.reset_system(reboot_mode ?
555 EFI_SUCCESS, 0, NULL);
556 reboot_type = BOOT_KBD;
560 port_cf9_safe = true;
565 u8 cf9 = inb(0xcf9) & ~6;
566 outb(cf9|2, 0xcf9); /* Request hard reset */
568 outb(cf9|6, 0xcf9); /* Actually do the reset */
571 reboot_type = BOOT_KBD;
577 void native_machine_shutdown(void)
579 /* Stop the cpus and apics */
582 /* The boot cpu is always logical cpu 0 */
583 int reboot_cpu_id = 0;
586 /* See if there has been given a command line override */
587 if ((reboot_cpu != -1) && (reboot_cpu < nr_cpu_ids) &&
588 cpu_online(reboot_cpu))
589 reboot_cpu_id = reboot_cpu;
592 /* Make certain the cpu I'm about to reboot on is online */
593 if (!cpu_online(reboot_cpu_id))
594 reboot_cpu_id = smp_processor_id();
596 /* Make certain I only run on the appropriate processor */
597 set_cpus_allowed_ptr(current, cpumask_of(reboot_cpu_id));
599 /* O.K Now that I'm on the appropriate processor,
600 * stop all of the others.
607 #ifdef CONFIG_X86_IO_APIC
611 #ifdef CONFIG_HPET_TIMER
616 pci_iommu_shutdown();
620 static void __machine_emergency_restart(int emergency)
622 reboot_emergency = emergency;
623 machine_ops.emergency_restart();
626 static void native_machine_restart(char *__unused)
628 printk("machine restart\n");
632 __machine_emergency_restart(0);
635 static void native_machine_halt(void)
637 /* stop other cpus and apics */
640 tboot_shutdown(TB_SHUTDOWN_HALT);
646 static void native_machine_power_off(void)
653 /* a fallback in case there is no PM info available */
654 tboot_shutdown(TB_SHUTDOWN_HALT);
657 struct machine_ops machine_ops = {
658 .power_off = native_machine_power_off,
659 .shutdown = native_machine_shutdown,
660 .emergency_restart = native_machine_emergency_restart,
661 .restart = native_machine_restart,
662 .halt = native_machine_halt,
664 .crash_shutdown = native_machine_crash_shutdown,
668 void machine_power_off(void)
670 machine_ops.power_off();
673 void machine_shutdown(void)
675 machine_ops.shutdown();
678 void machine_emergency_restart(void)
680 __machine_emergency_restart(1);
683 void machine_restart(char *cmd)
685 machine_ops.restart(cmd);
688 void machine_halt(void)
694 void machine_crash_shutdown(struct pt_regs *regs)
696 machine_ops.crash_shutdown(regs);
701 #if defined(CONFIG_SMP)
703 /* This keeps a track of which one is crashing cpu. */
704 static int crashing_cpu;
705 static nmi_shootdown_cb shootdown_callback;
707 static atomic_t waiting_for_crash_ipi;
709 static int crash_nmi_callback(struct notifier_block *self,
710 unsigned long val, void *data)
714 if (val != DIE_NMI_IPI)
717 cpu = raw_smp_processor_id();
719 /* Don't do anything if this handler is invoked on crashing cpu.
720 * Otherwise, system will completely hang. Crashing cpu can get
721 * an NMI if system was initially booted with nmi_watchdog parameter.
723 if (cpu == crashing_cpu)
727 shootdown_callback(cpu, (struct die_args *)data);
729 atomic_dec(&waiting_for_crash_ipi);
730 /* Assume hlt works */
738 static void smp_send_nmi_allbutself(void)
740 apic->send_IPI_allbutself(NMI_VECTOR);
743 static struct notifier_block crash_nmi_nb = {
744 .notifier_call = crash_nmi_callback,
747 /* Halt all other CPUs, calling the specified function on each of them
749 * This function can be used to halt all other CPUs on crash
750 * or emergency reboot time. The function passed as parameter
751 * will be called inside a NMI handler on all CPUs.
753 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
758 /* Make a note of crashing cpu. Will be used in NMI callback.*/
759 crashing_cpu = safe_smp_processor_id();
761 shootdown_callback = callback;
763 atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
764 /* Would it be better to replace the trap vector here? */
765 if (register_die_notifier(&crash_nmi_nb))
766 return; /* return what? */
767 /* Ensure the new callback function is set before sending
772 smp_send_nmi_allbutself();
774 msecs = 1000; /* Wait at most a second for the other cpus to stop */
775 while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
780 /* Leave the nmi callback set */
782 #else /* !CONFIG_SMP */
783 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
785 /* No other CPUs to shoot down */