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"),
219 { /* Handle problems with rebooting on Dell XPS710 */
220 .callback = set_bios_reboot,
221 .ident = "Dell XPS710",
223 DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
224 DMI_MATCH(DMI_PRODUCT_NAME, "Dell XPS710"),
230 static int __init reboot_init(void)
232 dmi_check_system(reboot_dmi_table);
235 core_initcall(reboot_init);
237 /* The following code and data reboots the machine by switching to real
238 mode and jumping to the BIOS reset entry point, as if the CPU has
239 really been reset. The previous version asked the keyboard
240 controller to pulse the CPU reset line, which is more thorough, but
241 doesn't work with at least one type of 486 motherboard. It is easy
242 to stop this code working; hence the copious comments. */
243 static const unsigned long long
244 real_mode_gdt_entries [3] =
246 0x0000000000000000ULL, /* Null descriptor */
247 0x00009b000000ffffULL, /* 16-bit real-mode 64k code at 0x00000000 */
248 0x000093000100ffffULL /* 16-bit real-mode 64k data at 0x00000100 */
251 static const struct desc_ptr
252 real_mode_gdt = { sizeof (real_mode_gdt_entries) - 1, (long)real_mode_gdt_entries },
253 real_mode_idt = { 0x3ff, 0 };
255 /* This is 16-bit protected mode code to disable paging and the cache,
256 switch to real mode and jump to the BIOS reset code.
258 The instruction that switches to real mode by writing to CR0 must be
259 followed immediately by a far jump instruction, which set CS to a
260 valid value for real mode, and flushes the prefetch queue to avoid
261 running instructions that have already been decoded in protected
264 Clears all the flags except ET, especially PG (paging), PE
265 (protected-mode enable) and TS (task switch for coprocessor state
266 save). Flushes the TLB after paging has been disabled. Sets CD and
267 NW, to disable the cache on a 486, and invalidates the cache. This
268 is more like the state of a 486 after reset. I don't know if
269 something else should be done for other chips.
271 More could be done here to set up the registers as if a CPU reset had
272 occurred; hopefully real BIOSs don't assume much. */
273 static const unsigned char real_mode_switch [] =
275 0x66, 0x0f, 0x20, 0xc0, /* movl %cr0,%eax */
276 0x66, 0x83, 0xe0, 0x11, /* andl $0x00000011,%eax */
277 0x66, 0x0d, 0x00, 0x00, 0x00, 0x60, /* orl $0x60000000,%eax */
278 0x66, 0x0f, 0x22, 0xc0, /* movl %eax,%cr0 */
279 0x66, 0x0f, 0x22, 0xd8, /* movl %eax,%cr3 */
280 0x66, 0x0f, 0x20, 0xc3, /* movl %cr0,%ebx */
281 0x66, 0x81, 0xe3, 0x00, 0x00, 0x00, 0x60, /* andl $0x60000000,%ebx */
282 0x74, 0x02, /* jz f */
283 0x0f, 0x09, /* wbinvd */
284 0x24, 0x10, /* f: andb $0x10,al */
285 0x66, 0x0f, 0x22, 0xc0 /* movl %eax,%cr0 */
287 static const unsigned char jump_to_bios [] =
289 0xea, 0x00, 0x00, 0xff, 0xff /* ljmp $0xffff,$0x0000 */
293 * Switch to real mode and then execute the code
294 * specified by the code and length parameters.
295 * We assume that length will aways be less that 100!
297 void machine_real_restart(const unsigned char *code, int length)
301 /* Write zero to CMOS register number 0x0f, which the BIOS POST
302 routine will recognize as telling it to do a proper reboot. (Well
303 that's what this book in front of me says -- it may only apply to
304 the Phoenix BIOS though, it's not clear). At the same time,
305 disable NMIs by setting the top bit in the CMOS address register,
306 as we're about to do peculiar things to the CPU. I'm not sure if
307 `outb_p' is needed instead of just `outb'. Use it to be on the
308 safe side. (Yes, CMOS_WRITE does outb_p's. - Paul G.)
310 spin_lock(&rtc_lock);
311 CMOS_WRITE(0x00, 0x8f);
312 spin_unlock(&rtc_lock);
314 /* Remap the kernel at virtual address zero, as well as offset zero
315 from the kernel segment. This assumes the kernel segment starts at
316 virtual address PAGE_OFFSET. */
317 memcpy(swapper_pg_dir, swapper_pg_dir + KERNEL_PGD_BOUNDARY,
318 sizeof(swapper_pg_dir [0]) * KERNEL_PGD_PTRS);
321 * Use `swapper_pg_dir' as our page directory.
323 load_cr3(swapper_pg_dir);
325 /* Write 0x1234 to absolute memory location 0x472. The BIOS reads
326 this on booting to tell it to "Bypass memory test (also warm
327 boot)". This seems like a fairly standard thing that gets set by
328 REBOOT.COM programs, and the previous reset routine did this
330 *((unsigned short *)0x472) = reboot_mode;
332 /* For the switch to real mode, copy some code to low memory. It has
333 to be in the first 64k because it is running in 16-bit mode, and it
334 has to have the same physical and virtual address, because it turns
335 off paging. Copy it near the end of the first page, out of the way
336 of BIOS variables. */
337 memcpy((void *)(0x1000 - sizeof(real_mode_switch) - 100),
338 real_mode_switch, sizeof (real_mode_switch));
339 memcpy((void *)(0x1000 - 100), code, length);
341 /* Set up the IDT for real mode. */
342 load_idt(&real_mode_idt);
344 /* Set up a GDT from which we can load segment descriptors for real
345 mode. The GDT is not used in real mode; it is just needed here to
346 prepare the descriptors. */
347 load_gdt(&real_mode_gdt);
349 /* Load the data segment registers, and thus the descriptors ready for
350 real mode. The base address of each segment is 0x100, 16 times the
351 selector value being loaded here. This is so that the segment
352 registers don't have to be reloaded after switching to real mode:
353 the values are consistent for real mode operation already. */
354 __asm__ __volatile__ ("movl $0x0010,%%eax\n"
355 "\tmovl %%eax,%%ds\n"
356 "\tmovl %%eax,%%es\n"
357 "\tmovl %%eax,%%fs\n"
358 "\tmovl %%eax,%%gs\n"
359 "\tmovl %%eax,%%ss" : : : "eax");
361 /* Jump to the 16-bit code that we copied earlier. It disables paging
362 and the cache, switches to real mode, and jumps to the BIOS reset
364 __asm__ __volatile__ ("ljmp $0x0008,%0"
366 : "i" ((void *)(0x1000 - sizeof (real_mode_switch) - 100)));
368 #ifdef CONFIG_APM_MODULE
369 EXPORT_SYMBOL(machine_real_restart);
372 #endif /* CONFIG_X86_32 */
374 static inline void kb_wait(void)
378 for (i = 0; i < 0x10000; i++) {
379 if ((inb(0x64) & 0x02) == 0)
385 static void vmxoff_nmi(int cpu, struct die_args *args)
387 cpu_emergency_vmxoff();
390 /* Use NMIs as IPIs to tell all CPUs to disable virtualization
392 static void emergency_vmx_disable_all(void)
394 /* Just make sure we won't change CPUs while doing this */
397 /* We need to disable VMX on all CPUs before rebooting, otherwise
398 * we risk hanging up the machine, because the CPU ignore INIT
399 * signals when VMX is enabled.
401 * We can't take any locks and we may be on an inconsistent
402 * state, so we use NMIs as IPIs to tell the other CPUs to disable
405 * For safety, we will avoid running the nmi_shootdown_cpus()
406 * stuff unnecessarily, but we don't have a way to check
407 * if other CPUs have VMX enabled. So we will call it only if the
408 * CPU we are running on has VMX enabled.
410 * We will miss cases where VMX is not enabled on all CPUs. This
411 * shouldn't do much harm because KVM always enable VMX on all
412 * CPUs anyway. But we can miss it on the small window where KVM
413 * is still enabling VMX.
415 if (cpu_has_vmx() && cpu_vmx_enabled()) {
416 /* Disable VMX on this CPU.
420 /* Halt and disable VMX on the other CPUs */
421 nmi_shootdown_cpus(vmxoff_nmi);
427 void __attribute__((weak)) mach_reboot_fixups(void)
431 static void native_machine_emergency_restart(void)
435 if (reboot_emergency)
436 emergency_vmx_disable_all();
438 /* Tell the BIOS if we want cold or warm reboot */
439 *((unsigned short *)__va(0x472)) = reboot_mode;
442 /* Could also try the reset bit in the Hammer NB */
443 switch (reboot_type) {
445 mach_reboot_fixups(); /* for board specific fixups */
447 for (i = 0; i < 10; i++) {
450 outb(0xfe, 0x64); /* pulse reset low */
456 __asm__ __volatile__("int3");
458 reboot_type = BOOT_KBD;
463 machine_real_restart(jump_to_bios, sizeof(jump_to_bios));
465 reboot_type = BOOT_KBD;
471 reboot_type = BOOT_KBD;
476 efi.reset_system(reboot_mode ?
479 EFI_SUCCESS, 0, NULL);
480 reboot_type = BOOT_KBD;
484 port_cf9_safe = true;
489 u8 cf9 = inb(0xcf9) & ~6;
490 outb(cf9|2, 0xcf9); /* Request hard reset */
492 outb(cf9|6, 0xcf9); /* Actually do the reset */
495 reboot_type = BOOT_KBD;
501 void native_machine_shutdown(void)
503 /* Stop the cpus and apics */
506 /* The boot cpu is always logical cpu 0 */
507 int reboot_cpu_id = 0;
510 /* See if there has been given a command line override */
511 if ((reboot_cpu != -1) && (reboot_cpu < nr_cpu_ids) &&
512 cpu_online(reboot_cpu))
513 reboot_cpu_id = reboot_cpu;
516 /* Make certain the cpu I'm about to reboot on is online */
517 if (!cpu_online(reboot_cpu_id))
518 reboot_cpu_id = smp_processor_id();
520 /* Make certain I only run on the appropriate processor */
521 set_cpus_allowed_ptr(current, cpumask_of(reboot_cpu_id));
523 /* O.K Now that I'm on the appropriate processor,
524 * stop all of the others.
531 #ifdef CONFIG_X86_IO_APIC
535 #ifdef CONFIG_HPET_TIMER
540 pci_iommu_shutdown();
544 static void __machine_emergency_restart(int emergency)
546 reboot_emergency = emergency;
547 machine_ops.emergency_restart();
550 static void native_machine_restart(char *__unused)
552 printk("machine restart\n");
556 __machine_emergency_restart(0);
559 static void native_machine_halt(void)
561 /* stop other cpus and apics */
568 static void native_machine_power_off(void)
577 struct machine_ops machine_ops = {
578 .power_off = native_machine_power_off,
579 .shutdown = native_machine_shutdown,
580 .emergency_restart = native_machine_emergency_restart,
581 .restart = native_machine_restart,
582 .halt = native_machine_halt,
584 .crash_shutdown = native_machine_crash_shutdown,
588 void machine_power_off(void)
590 machine_ops.power_off();
593 void machine_shutdown(void)
595 machine_ops.shutdown();
598 void machine_emergency_restart(void)
600 __machine_emergency_restart(1);
603 void machine_restart(char *cmd)
605 machine_ops.restart(cmd);
608 void machine_halt(void)
614 void machine_crash_shutdown(struct pt_regs *regs)
616 machine_ops.crash_shutdown(regs);
621 #if defined(CONFIG_SMP)
623 /* This keeps a track of which one is crashing cpu. */
624 static int crashing_cpu;
625 static nmi_shootdown_cb shootdown_callback;
627 static atomic_t waiting_for_crash_ipi;
629 static int crash_nmi_callback(struct notifier_block *self,
630 unsigned long val, void *data)
634 if (val != DIE_NMI_IPI)
637 cpu = raw_smp_processor_id();
639 /* Don't do anything if this handler is invoked on crashing cpu.
640 * Otherwise, system will completely hang. Crashing cpu can get
641 * an NMI if system was initially booted with nmi_watchdog parameter.
643 if (cpu == crashing_cpu)
647 shootdown_callback(cpu, (struct die_args *)data);
649 atomic_dec(&waiting_for_crash_ipi);
650 /* Assume hlt works */
658 static void smp_send_nmi_allbutself(void)
660 apic->send_IPI_allbutself(NMI_VECTOR);
663 static struct notifier_block crash_nmi_nb = {
664 .notifier_call = crash_nmi_callback,
667 /* Halt all other CPUs, calling the specified function on each of them
669 * This function can be used to halt all other CPUs on crash
670 * or emergency reboot time. The function passed as parameter
671 * will be called inside a NMI handler on all CPUs.
673 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
678 /* Make a note of crashing cpu. Will be used in NMI callback.*/
679 crashing_cpu = safe_smp_processor_id();
681 shootdown_callback = callback;
683 atomic_set(&waiting_for_crash_ipi, num_online_cpus() - 1);
684 /* Would it be better to replace the trap vector here? */
685 if (register_die_notifier(&crash_nmi_nb))
686 return; /* return what? */
687 /* Ensure the new callback function is set before sending
692 smp_send_nmi_allbutself();
694 msecs = 1000; /* Wait at most a second for the other cpus to stop */
695 while ((atomic_read(&waiting_for_crash_ipi) > 0) && msecs) {
700 /* Leave the nmi callback set */
702 #else /* !CONFIG_SMP */
703 void nmi_shootdown_cpus(nmi_shootdown_cb callback)
705 /* No other CPUs to shoot down */