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 2400's */
196 .callback = set_bios_reboot,
197 .ident = "Dell PowerEdge 2400",
199 DMI_MATCH(DMI_SYS_VENDOR, "Dell Computer Corporation"),
200 DMI_MATCH(DMI_PRODUCT_NAME, "PowerEdge 2400"),
203 { /* Handle problems with rebooting on Dell T5400's */
204 .callback = set_bios_reboot,
205 .ident = "Dell Precision T5400",
207 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
208 DMI_MATCH(DMI_PRODUCT_NAME, "Precision WorkStation T5400"),
211 { /* Handle problems with rebooting on HP laptops */
212 .callback = set_bios_reboot,
213 .ident = "HP Compaq Laptop",
215 DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"),
216 DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq"),
222 static int __init reboot_init(void)
224 dmi_check_system(reboot_dmi_table);
227 core_initcall(reboot_init);
229 /* The following code and data reboots the machine by switching to real
230 mode and jumping to the BIOS reset entry point, as if the CPU has
231 really been reset. The previous version asked the keyboard
232 controller to pulse the CPU reset line, which is more thorough, but
233 doesn't work with at least one type of 486 motherboard. It is easy
234 to stop this code working; hence the copious comments. */
235 static const unsigned long long
236 real_mode_gdt_entries [3] =
238 0x0000000000000000ULL, /* Null descriptor */
239 0x00009b000000ffffULL, /* 16-bit real-mode 64k code at 0x00000000 */
240 0x000093000100ffffULL /* 16-bit real-mode 64k data at 0x00000100 */
243 static const struct desc_ptr
244 real_mode_gdt = { sizeof (real_mode_gdt_entries) - 1, (long)real_mode_gdt_entries },
245 real_mode_idt = { 0x3ff, 0 };
247 /* This is 16-bit protected mode code to disable paging and the cache,
248 switch to real mode and jump to the BIOS reset code.
250 The instruction that switches to real mode by writing to CR0 must be
251 followed immediately by a far jump instruction, which set CS to a
252 valid value for real mode, and flushes the prefetch queue to avoid
253 running instructions that have already been decoded in protected
256 Clears all the flags except ET, especially PG (paging), PE
257 (protected-mode enable) and TS (task switch for coprocessor state
258 save). Flushes the TLB after paging has been disabled. Sets CD and
259 NW, to disable the cache on a 486, and invalidates the cache. This
260 is more like the state of a 486 after reset. I don't know if
261 something else should be done for other chips.
263 More could be done here to set up the registers as if a CPU reset had
264 occurred; hopefully real BIOSs don't assume much. */
265 static const unsigned char real_mode_switch [] =
267 0x66, 0x0f, 0x20, 0xc0, /* movl %cr0,%eax */
268 0x66, 0x83, 0xe0, 0x11, /* andl $0x00000011,%eax */
269 0x66, 0x0d, 0x00, 0x00, 0x00, 0x60, /* orl $0x60000000,%eax */
270 0x66, 0x0f, 0x22, 0xc0, /* movl %eax,%cr0 */
271 0x66, 0x0f, 0x22, 0xd8, /* movl %eax,%cr3 */
272 0x66, 0x0f, 0x20, 0xc3, /* movl %cr0,%ebx */
273 0x66, 0x81, 0xe3, 0x00, 0x00, 0x00, 0x60, /* andl $0x60000000,%ebx */
274 0x74, 0x02, /* jz f */
275 0x0f, 0x09, /* wbinvd */
276 0x24, 0x10, /* f: andb $0x10,al */
277 0x66, 0x0f, 0x22, 0xc0 /* movl %eax,%cr0 */
279 static const unsigned char jump_to_bios [] =
281 0xea, 0x00, 0x00, 0xff, 0xff /* ljmp $0xffff,$0x0000 */
285 * Switch to real mode and then execute the code
286 * specified by the code and length parameters.
287 * We assume that length will aways be less that 100!
289 void machine_real_restart(const unsigned char *code, int length)
293 /* Write zero to CMOS register number 0x0f, which the BIOS POST
294 routine will recognize as telling it to do a proper reboot. (Well
295 that's what this book in front of me says -- it may only apply to
296 the Phoenix BIOS though, it's not clear). At the same time,
297 disable NMIs by setting the top bit in the CMOS address register,
298 as we're about to do peculiar things to the CPU. I'm not sure if
299 `outb_p' is needed instead of just `outb'. Use it to be on the
300 safe side. (Yes, CMOS_WRITE does outb_p's. - Paul G.)
302 spin_lock(&rtc_lock);
303 CMOS_WRITE(0x00, 0x8f);
304 spin_unlock(&rtc_lock);
306 /* Remap the kernel at virtual address zero, as well as offset zero
307 from the kernel segment. This assumes the kernel segment starts at
308 virtual address PAGE_OFFSET. */
309 memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
310 sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);
313 * Use `swapper_pg_dir' as our page directory.
315 load_cr3(swapper_pg_dir);
317 /* Write 0x1234 to absolute memory location 0x472. The BIOS reads
318 this on booting to tell it to "Bypass memory test (also warm
319 boot)". This seems like a fairly standard thing that gets set by
320 REBOOT.COM programs, and the previous reset routine did this
322 *((unsigned short *)0x472) = reboot_mode;
324 /* For the switch to real mode, copy some code to low memory. It has
325 to be in the first 64k because it is running in 16-bit mode, and it
326 has to have the same physical and virtual address, because it turns
327 off paging. Copy it near the end of the first page, out of the way
328 of BIOS variables. */
329 memcpy((void *)(0x1000 - sizeof(real_mode_switch) - 100),
330 real_mode_switch, sizeof (real_mode_switch));
331 memcpy((void *)(0x1000 - 100), code, length);
333 /* Set up the IDT for real mode. */
334 load_idt(&real_mode_idt);
336 /* Set up a GDT from which we can load segment descriptors for real
337 mode. The GDT is not used in real mode; it is just needed here to
338 prepare the descriptors. */
339 load_gdt(&real_mode_gdt);
341 /* Load the data segment registers, and thus the descriptors ready for
342 real mode. The base address of each segment is 0x100, 16 times the
343 selector value being loaded here. This is so that the segment
344 registers don't have to be reloaded after switching to real mode:
345 the values are consistent for real mode operation already. */
346 __asm__ __volatile__ ("movl $0x0010,%%eax\n"
347 "\tmovl %%eax,%%ds\n"
348 "\tmovl %%eax,%%es\n"
349 "\tmovl %%eax,%%fs\n"
350 "\tmovl %%eax,%%gs\n"
351 "\tmovl %%eax,%%ss" : : : "eax");
353 /* Jump to the 16-bit code that we copied earlier. It disables paging
354 and the cache, switches to real mode, and jumps to the BIOS reset
356 __asm__ __volatile__ ("ljmp $0x0008,%0"
358 : "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
360 #ifdef CONFIG_APM_MODULE
361 EXPORT_SYMBOL(machine_real_restart);
364 #endif /* CONFIG_X86_32 */
366 static inline void kb_wait(void)
370 for (i = 0; i < 0x10000; i++) {
371 if ((inb(0x64) & 0x02) == 0)
377 static void vmxoff_nmi(int cpu, struct die_args *args)
379 cpu_emergency_vmxoff();
382 /* Use NMIs as IPIs to tell all CPUs to disable virtualization
384 static void emergency_vmx_disable_all(void)
386 /* Just make sure we won't change CPUs while doing this */
389 /* We need to disable VMX on all CPUs before rebooting, otherwise
390 * we risk hanging up the machine, because the CPU ignore INIT
391 * signals when VMX is enabled.
393 * We can't take any locks and we may be on an inconsistent
394 * state, so we use NMIs as IPIs to tell the other CPUs to disable
397 * For safety, we will avoid running the nmi_shootdown_cpus()
398 * stuff unnecessarily, but we don't have a way to check
399 * if other CPUs have VMX enabled. So we will call it only if the
400 * CPU we are running on has VMX enabled.
402 * We will miss cases where VMX is not enabled on all CPUs. This
403 * shouldn't do much harm because KVM always enable VMX on all
404 * CPUs anyway. But we can miss it on the small window where KVM
405 * is still enabling VMX.
407 if (cpu_has_vmx() && cpu_vmx_enabled()) {
408 /* Disable VMX on this CPU.
412 /* Halt and disable VMX on the other CPUs */
413 nmi_shootdown_cpus(vmxoff_nmi);
419 void __attribute__((weak)) mach_reboot_fixups(void)
423 static void native_machine_emergency_restart(void)
427 if (reboot_emergency)
428 emergency_vmx_disable_all();
430 /* Tell the BIOS if we want cold or warm reboot */
431 *((unsigned short *)__va(0x472)) = reboot_mode;
434 /* Could also try the reset bit in the Hammer NB */
435 switch (reboot_type) {
437 mach_reboot_fixups(); /* for board specific fixups */
439 for (i = 0; i < 10; i++) {
442 outb(0xfe, 0x64); /* pulse reset low */
448 __asm__ __volatile__("int3");
450 reboot_type = BOOT_KBD;
455 machine_real_restart(jump_to_bios, sizeof(jump_to_bios));
457 reboot_type = BOOT_KBD;
463 reboot_type = BOOT_KBD;
468 efi.reset_system(reboot_mode ?
471 EFI_SUCCESS, 0, NULL);
472 reboot_type = BOOT_KBD;
476 port_cf9_safe = true;
481 u8 cf9 = inb(0xcf9) & ~6;
482 outb(cf9|2, 0xcf9); /* Request hard reset */
484 outb(cf9|6, 0xcf9); /* Actually do the reset */
487 reboot_type = BOOT_KBD;
493 void native_machine_shutdown(void)
495 /* Stop the cpus and apics */
498 /* The boot cpu is always logical cpu 0 */
499 int reboot_cpu_id = 0;
502 /* See if there has been given a command line override */
503 if ((reboot_cpu != -1) && (reboot_cpu < nr_cpu_ids) &&
504 cpu_online(reboot_cpu))
505 reboot_cpu_id = reboot_cpu;
508 /* Make certain the cpu I'm about to reboot on is online */
509 if (!cpu_online(reboot_cpu_id))
510 reboot_cpu_id = smp_processor_id();
512 /* Make certain I only run on the appropriate processor */
513 set_cpus_allowed_ptr(current, cpumask_of(reboot_cpu_id));
515 /* O.K Now that I'm on the appropriate processor,
516 * stop all of the others.
523 #ifdef CONFIG_X86_IO_APIC
527 #ifdef CONFIG_HPET_TIMER
532 pci_iommu_shutdown();
536 static void __machine_emergency_restart(int emergency)
538 reboot_emergency = emergency;
539 machine_ops.emergency_restart();
542 static void native_machine_restart(char *__unused)
544 printk("machine restart\n");
548 __machine_emergency_restart(0);
551 static void native_machine_halt(void)
553 /* stop other cpus and apics */
560 static void native_machine_power_off(void)
569 struct machine_ops machine_ops = {
570 .power_off = native_machine_power_off,
571 .shutdown = native_machine_shutdown,
572 .emergency_restart = native_machine_emergency_restart,
573 .restart = native_machine_restart,
574 .halt = native_machine_halt,
576 .crash_shutdown = native_machine_crash_shutdown,
580 void machine_power_off(void)
582 machine_ops.power_off();
585 void machine_shutdown(void)
587 machine_ops.shutdown();
590 void machine_emergency_restart(void)
592 __machine_emergency_restart(1);
595 void machine_restart(char *cmd)
597 machine_ops.restart(cmd);
600 void machine_halt(void)
606 void machine_crash_shutdown(struct pt_regs *regs)
608 machine_ops.crash_shutdown(regs);
613 #if defined(CONFIG_SMP)
615 /* This keeps a track of which one is crashing cpu. */
616 static int crashing_cpu;
617 static nmi_shootdown_cb shootdown_callback;
619 static atomic_t waiting_for_crash_ipi;
621 static int crash_nmi_callback(struct notifier_block *self,
622 unsigned long val, void *data)
626 if (val != DIE_NMI_IPI)
629 cpu = raw_smp_processor_id();
631 /* Don't do anything if this handler is invoked on crashing cpu.
632 * Otherwise, system will completely hang. Crashing cpu can get
633 * an NMI if system was initially booted with nmi_watchdog parameter.
635 if (cpu == crashing_cpu)
639 shootdown_callback(cpu, (struct die_args *)data);
641 atomic_dec(&waiting_for_crash_ipi);
642 /* Assume hlt works */
650 static void smp_send_nmi_allbutself(void)
652 apic->send_IPI_allbutself(NMI_VECTOR);
655 static struct notifier_block crash_nmi_nb = {
656 .notifier_call = crash_nmi_callback,
659 /* Halt all other CPUs, calling the specified function on each of them
661 * This function can be used to halt all other CPUs on crash
662 * or emergency reboot time. The function passed as parameter
663 * will be called inside a NMI handler on all CPUs.
665 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
670 /* Make a note of crashing cpu. Will be used in NMI callback.*/
671 crashing_cpu = safe_smp_processor_id();
673 shootdown_callback = callback;
675 atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
676 /* Would it be better to replace the trap vector here? */
677 if (register_die_notifier(&crash_nmi_nb))
678 return; /* return what? */
679 /* Ensure the new callback function is set before sending
684 smp_send_nmi_allbutself();
686 msecs = 1000; /* Wait at most a second for the other cpus to stop */
687 while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
692 /* Leave the nmi callback set */
694 #else /* !CONFIG_SMP */
695 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
697 /* No other CPUs to shoot down */