1 #include <linux/module.h>
2 #include <linux/reboot.h>
3 #include <linux/init.h>
6 #include <acpi/reboot.h>
11 #include <asm/pgtable.h>
12 #include <asm/proto.h>
13 #include <asm/reboot_fixups.h>
14 #include <asm/reboot.h>
15 #include <asm/pci_x86.h>
16 #include <asm/virtext.h>
20 # include <linux/dmi.h>
21 # include <linux/ctype.h>
22 # include <linux/mc146818rtc.h>
24 # include <asm/iommu.h>
28 * Power off function, if any
30 void (*pm_power_off)(void);
31 EXPORT_SYMBOL(pm_power_off);
33 static const struct desc_ptr no_idt = {};
34 static int reboot_mode;
35 enum reboot_type reboot_type = BOOT_KBD;
38 #if defined(CONFIG_X86_32) && defined(CONFIG_SMP)
39 static int reboot_cpu = -1;
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
46 static int reboot_emergency;
48 /* This is set by the PCI code if either type 1 or type 2 PCI is detected */
49 bool port_cf9_safe = false;
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.
63 static int __init reboot_setup(char *str)
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');
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 */
87 #endif /* CONFIG_SMP */
104 str = strchr(str, ',');
113 __setup("reboot=", reboot_setup);
118 * Reboot options and system auto-detection code provided by
119 * Dell Inc. so their systems "just work". :-)
123 * Some machines require the "reboot=b" commandline option,
124 * this quirk makes that automatic.
126 static int __init set_bios_reboot(const struct dmi_system_id *d)
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);
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",
140 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
141 DMI_MATCH(DMI_PRODUCT_NAME, "Dell DM061"),
144 { /* Handle problems with rebooting on Dell 1300's */
145 .callback = set_bios_reboot,
146 .ident = "Dell PowerEdge 1300",
148 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
149 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 1300/"),
152 { /* Handle problems with rebooting on Dell 300's */
153 .callback = set_bios_reboot,
154 .ident = "Dell PowerEdge 300",
156 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
157 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 300/"),
160 { /* Handle problems with rebooting on Dell Optiplex 745's SFF*/
161 .callback = set_bios_reboot,
162 .ident = "Dell OptiPlex 745",
164 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
165 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
168 { /* Handle problems with rebooting on Dell Optiplex 745's DFF*/
169 .callback = set_bios_reboot,
170 .ident = "Dell OptiPlex 745",
172 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
173 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
174 DMI_MATCH(DMI_BOARD_NAME, "0MM599"),
177 { /* Handle problems with rebooting on Dell Optiplex 745 with 0KW626 */
178 .callback = set_bios_reboot,
179 .ident = "Dell OptiPlex 745",
181 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
182 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 745"),
183 DMI_MATCH(DMI_BOARD_NAME, "0KW626"),
186 { /* Handle problems with rebooting on Dell Optiplex 330 with 0KP561 */
187 .callback = set_bios_reboot,
188 .ident = "Dell OptiPlex 330",
190 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
191 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 330"),
192 DMI_MATCH(DMI_BOARD_NAME, "0KP561"),
195 { /* Handle problems with rebooting on Dell Optiplex 360 with 0T656F */
196 .callback = set_bios_reboot,
197 .ident = "Dell OptiPlex 360",
199 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
200 DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 360"),
201 DMI_MATCH(DMI_BOARD_NAME, "0T656F"),
204 { /* Handle problems with rebooting on Dell 2400's */
205 .callback = set_bios_reboot,
206 .ident = "Dell PowerEdge 2400",
208 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
209 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"),
212 { /* Handle problems with rebooting on Dell T5400's */
213 .callback = set_bios_reboot,
214 .ident = "Dell Precision T5400",
216 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
217 DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"),
220 { /* Handle problems with rebooting on HP laptops */
221 .callback = set_bios_reboot,
222 .ident = "HP Compaq Laptop",
224 DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
225 DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq"),
228 { /* Handle problems with rebooting on Dell XPS710 */
229 .callback = set_bios_reboot,
230 .ident = "Dell XPS710",
232 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
233 DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
236 { /* Handle problems with rebooting on Dell DXP061 */
237 .callback = set_bios_reboot,
238 .ident = "Dell DXP061",
240 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
241 DMI_MATCH(DMI_PRODUCT_NAME, "Dell DXP061"),
244 { /* Handle problems with rebooting on Sony VGN-Z540N */
245 .callback = set_bios_reboot,
246 .ident = "Sony VGN-Z540N",
248 DMI_MATCH(DMI_SYS_VENDOR, "Sony Corporation"),
249 DMI_MATCH(DMI_PRODUCT_NAME, "VGN-Z540N"),
252 { /* Handle problems with rebooting on CompuLab SBC-FITPC2 */
253 .callback = set_bios_reboot,
254 .ident = "CompuLab SBC-FITPC2",
256 DMI_MATCH(DMI_SYS_VENDOR, "CompuLab"),
257 DMI_MATCH(DMI_PRODUCT_NAME, "SBC-FITPC2"),
263 static int __init reboot_init(void)
265 dmi_check_system(reboot_dmi_table);
268 core_initcall(reboot_init);
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] =
279 0x0000000000000000ULL, /* Null descriptor */
280 0x00009b000000ffffULL, /* 16-bit real-mode 64k code at 0x00000000 */
281 0x000093000100ffffULL /* 16-bit real-mode 64k data at 0x00000100 */
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 };
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.
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
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.
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 [] =
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 */
320 static const unsigned char jump_to_bios [] =
322 0xea, 0x00, 0x00, 0xff, 0xff /* ljmp $0xffff,$0x0000 */
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!
330 void machine_real_restart(const unsigned char *code, int length)
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.)
343 spin_lock(&rtc_lock);
344 CMOS_WRITE(0x00, 0x8f);
345 spin_unlock(&rtc_lock);
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);
354 * Use `swapper_pg_dir' as our page directory.
356 load_cr3(swapper_pg_dir);
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
363 *((unsigned short *)0x472) = reboot_mode;
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);
374 /* Set up the IDT for real mode. */
375 load_idt(&real_mode_idt);
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);
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");
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
397 __asm__ __volatile__ ("ljmp $0x0008,%0"
399 : "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
401 #ifdef CONFIG_APM_MODULE
402 EXPORT_SYMBOL(machine_real_restart);
405 #endif /* CONFIG_X86_32 */
407 static inline void kb_wait(void)
411 for (i = 0; i < 0x10000; i++) {
412 if ((inb(0x64) & 0x02) == 0)
418 static void vmxoff_nmi(int cpu, struct die_args *args)
420 cpu_emergency_vmxoff();
423 /* Use NMIs as IPIs to tell all CPUs to disable virtualization
425 static void emergency_vmx_disable_all(void)
427 /* Just make sure we won't change CPUs while doing this */
430 /* We need to disable VMX on all CPUs before rebooting, otherwise
431 * we risk hanging up the machine, because the CPU ignore INIT
432 * signals when VMX is enabled.
434 * We can't take any locks and we may be on an inconsistent
435 * state, so we use NMIs as IPIs to tell the other CPUs to disable
438 * For safety, we will avoid running the nmi_shootdown_cpus()
439 * stuff unnecessarily, but we don't have a way to check
440 * if other CPUs have VMX enabled. So we will call it only if the
441 * CPU we are running on has VMX enabled.
443 * We will miss cases where VMX is not enabled on all CPUs. This
444 * shouldn't do much harm because KVM always enable VMX on all
445 * CPUs anyway. But we can miss it on the small window where KVM
446 * is still enabling VMX.
448 if (cpu_has_vmx() && cpu_vmx_enabled()) {
449 /* Disable VMX on this CPU.
453 /* Halt and disable VMX on the other CPUs */
454 nmi_shootdown_cpus(vmxoff_nmi);
460 void __attribute__((weak)) mach_reboot_fixups(void)
464 static void native_machine_emergency_restart(void)
468 if (reboot_emergency)
469 emergency_vmx_disable_all();
471 /* Tell the BIOS if we want cold or warm reboot */
472 *((unsigned short *)__va(0x472)) = reboot_mode;
475 /* Could also try the reset bit in the Hammer NB */
476 switch (reboot_type) {
478 mach_reboot_fixups(); /* for board specific fixups */
480 for (i = 0; i < 10; i++) {
483 outb(0xfe, 0x64); /* pulse reset low */
489 __asm__ __volatile__("int3");
491 reboot_type = BOOT_KBD;
496 machine_real_restart(jump_to_bios, sizeof(jump_to_bios));
498 reboot_type = BOOT_KBD;
504 reboot_type = BOOT_KBD;
509 efi.reset_system(reboot_mode ?
512 EFI_SUCCESS, 0, NULL);
513 reboot_type = BOOT_KBD;
517 port_cf9_safe = true;
522 u8 cf9 = inb(0xcf9) & ~6;
523 outb(cf9|2, 0xcf9); /* Request hard reset */
525 outb(cf9|6, 0xcf9); /* Actually do the reset */
528 reboot_type = BOOT_KBD;
534 void native_machine_shutdown(void)
536 /* Stop the cpus and apics */
539 /* The boot cpu is always logical cpu 0 */
540 int reboot_cpu_id = 0;
543 /* See if there has been given a command line override */
544 if ((reboot_cpu != -1) && (reboot_cpu < nr_cpu_ids) &&
545 cpu_online(reboot_cpu))
546 reboot_cpu_id = reboot_cpu;
549 /* Make certain the cpu I'm about to reboot on is online */
550 if (!cpu_online(reboot_cpu_id))
551 reboot_cpu_id = smp_processor_id();
553 /* Make certain I only run on the appropriate processor */
554 set_cpus_allowed_ptr(current, cpumask_of(reboot_cpu_id));
556 /* O.K Now that I'm on the appropriate processor,
557 * stop all of the others.
564 #ifdef CONFIG_X86_IO_APIC
568 #ifdef CONFIG_HPET_TIMER
573 pci_iommu_shutdown();
577 static void __machine_emergency_restart(int emergency)
579 reboot_emergency = emergency;
580 machine_ops.emergency_restart();
583 static void native_machine_restart(char *__unused)
585 printk("machine restart\n");
589 __machine_emergency_restart(0);
592 static void native_machine_halt(void)
594 /* stop other cpus and apics */
601 static void native_machine_power_off(void)
610 struct machine_ops machine_ops = {
611 .power_off = native_machine_power_off,
612 .shutdown = native_machine_shutdown,
613 .emergency_restart = native_machine_emergency_restart,
614 .restart = native_machine_restart,
615 .halt = native_machine_halt,
617 .crash_shutdown = native_machine_crash_shutdown,
621 void machine_power_off(void)
623 machine_ops.power_off();
626 void machine_shutdown(void)
628 machine_ops.shutdown();
631 void machine_emergency_restart(void)
633 __machine_emergency_restart(1);
636 void machine_restart(char *cmd)
638 machine_ops.restart(cmd);
641 void machine_halt(void)
647 void machine_crash_shutdown(struct pt_regs *regs)
649 machine_ops.crash_shutdown(regs);
654 #if defined(CONFIG_SMP)
656 /* This keeps a track of which one is crashing cpu. */
657 static int crashing_cpu;
658 static nmi_shootdown_cb shootdown_callback;
660 static atomic_t waiting_for_crash_ipi;
662 static int crash_nmi_callback(struct notifier_block *self,
663 unsigned long val, void *data)
667 if (val != DIE_NMI_IPI)
670 cpu = raw_smp_processor_id();
672 /* Don't do anything if this handler is invoked on crashing cpu.
673 * Otherwise, system will completely hang. Crashing cpu can get
674 * an NMI if system was initially booted with nmi_watchdog parameter.
676 if (cpu == crashing_cpu)
680 shootdown_callback(cpu, (struct die_args *)data);
682 atomic_dec(&waiting_for_crash_ipi);
683 /* Assume hlt works */
691 static void smp_send_nmi_allbutself(void)
693 apic->send_IPI_allbutself(NMI_VECTOR);
696 static struct notifier_block crash_nmi_nb = {
697 .notifier_call = crash_nmi_callback,
700 /* Halt all other CPUs, calling the specified function on each of them
702 * This function can be used to halt all other CPUs on crash
703 * or emergency reboot time. The function passed as parameter
704 * will be called inside a NMI handler on all CPUs.
706 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
711 /* Make a note of crashing cpu. Will be used in NMI callback.*/
712 crashing_cpu = safe_smp_processor_id();
714 shootdown_callback = callback;
716 atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
717 /* Would it be better to replace the trap vector here? */
718 if (register_die_notifier(&crash_nmi_nb))
719 return; /* return what? */
720 /* Ensure the new callback function is set before sending
725 smp_send_nmi_allbutself();
727 msecs = 1000; /* Wait at most a second for the other cpus to stop */
728 while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
733 /* Leave the nmi callback set */
735 #else /* !CONFIG_SMP */
736 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
738 /* No other CPUs to shoot down */