xref: /linux/arch/x86/kernel/apic/apic.c (revision 7fc2cd2e4b398c57c9cf961cfea05eadbf34c05c)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *	Local APIC handling, local APIC timers
4  *
5  *	(c) 1999, 2000, 2009 Ingo Molnar <mingo@redhat.com>
6  *
7  *	Fixes
8  *	Maciej W. Rozycki	:	Bits for genuine 82489DX APICs;
9  *					thanks to Eric Gilmore
10  *					and Rolf G. Tews
11  *					for testing these extensively.
12  *	Maciej W. Rozycki	:	Various updates and fixes.
13  *	Mikael Pettersson	:	Power Management for UP-APIC.
14  *	Pavel Machek and
15  *	Mikael Pettersson	:	PM converted to driver model.
16  */
17 
18 #include <linux/perf_event.h>
19 #include <linux/kernel_stat.h>
20 #include <linux/mc146818rtc.h>
21 #include <linux/acpi_pmtmr.h>
22 #include <linux/bitmap.h>
23 #include <linux/clockchips.h>
24 #include <linux/interrupt.h>
25 #include <linux/memblock.h>
26 #include <linux/ftrace.h>
27 #include <linux/ioport.h>
28 #include <linux/export.h>
29 #include <linux/syscore_ops.h>
30 #include <linux/delay.h>
31 #include <linux/timex.h>
32 #include <linux/i8253.h>
33 #include <linux/dmar.h>
34 #include <linux/init.h>
35 #include <linux/cpu.h>
36 #include <linux/dmi.h>
37 #include <linux/smp.h>
38 #include <linux/mm.h>
39 #include <linux/kvm_types.h>
40 
41 #include <xen/xen.h>
42 
43 #include <asm/trace/irq_vectors.h>
44 #include <asm/irq_remapping.h>
45 #include <asm/pc-conf-reg.h>
46 #include <asm/perf_event.h>
47 #include <asm/x86_init.h>
48 #include <linux/atomic.h>
49 #include <asm/barrier.h>
50 #include <asm/mpspec.h>
51 #include <asm/i8259.h>
52 #include <asm/proto.h>
53 #include <asm/traps.h>
54 #include <asm/apic.h>
55 #include <asm/acpi.h>
56 #include <asm/io_apic.h>
57 #include <asm/desc.h>
58 #include <asm/hpet.h>
59 #include <asm/mtrr.h>
60 #include <asm/time.h>
61 #include <asm/smp.h>
62 #include <asm/mce.h>
63 #include <asm/msr.h>
64 #include <asm/tsc.h>
65 #include <asm/hypervisor.h>
66 #include <asm/cpu_device_id.h>
67 #include <asm/intel-family.h>
68 #include <asm/irq_regs.h>
69 #include <asm/cpu.h>
70 
71 #include "local.h"
72 
73 /* Processor that is doing the boot up */
74 u32 boot_cpu_physical_apicid __ro_after_init = BAD_APICID;
75 EXPORT_SYMBOL_GPL(boot_cpu_physical_apicid);
76 
77 u8 boot_cpu_apic_version __ro_after_init;
78 
79 /*
80  * This variable controls which CPUs receive external NMIs.  By default,
81  * external NMIs are delivered only to the BSP.
82  */
83 static int apic_extnmi __ro_after_init = APIC_EXTNMI_BSP;
84 
85 /*
86  * Hypervisor supports 15 bits of APIC ID in MSI Extended Destination ID
87  */
88 static bool virt_ext_dest_id __ro_after_init;
89 
90 /* For parallel bootup. */
91 unsigned long apic_mmio_base __ro_after_init;
92 
93 static inline bool apic_accessible(void)
94 {
95 	return x2apic_mode || apic_mmio_base;
96 }
97 
98 #ifdef CONFIG_X86_32
99 /* Local APIC was disabled by the BIOS and enabled by the kernel */
100 static int enabled_via_apicbase __ro_after_init;
101 
102 /*
103  * Handle interrupt mode configuration register (IMCR).
104  * This register controls whether the interrupt signals
105  * that reach the BSP come from the master PIC or from the
106  * local APIC. Before entering Symmetric I/O Mode, either
107  * the BIOS or the operating system must switch out of
108  * PIC Mode by changing the IMCR.
109  */
110 static inline void imcr_pic_to_apic(void)
111 {
112 	/* NMI and 8259 INTR go through APIC */
113 	pc_conf_set(PC_CONF_MPS_IMCR, 0x01);
114 }
115 
116 static inline void imcr_apic_to_pic(void)
117 {
118 	/* NMI and 8259 INTR go directly to BSP */
119 	pc_conf_set(PC_CONF_MPS_IMCR, 0x00);
120 }
121 #endif
122 
123 /*
124  * Knob to control our willingness to enable the local APIC.
125  *
126  * +1=force-enable
127  */
128 static int force_enable_local_apic __initdata;
129 
130 /*
131  * APIC command line parameters
132  */
133 static int __init parse_lapic(char *arg)
134 {
135 	if (IS_ENABLED(CONFIG_X86_32) && !arg)
136 		force_enable_local_apic = 1;
137 	else if (arg && !strncmp(arg, "notscdeadline", 13))
138 		setup_clear_cpu_cap(X86_FEATURE_TSC_DEADLINE_TIMER);
139 	return 0;
140 }
141 early_param("lapic", parse_lapic);
142 
143 #ifdef CONFIG_X86_64
144 static int apic_calibrate_pmtmr __initdata;
145 static __init int setup_apicpmtimer(char *s)
146 {
147 	apic_calibrate_pmtmr = 1;
148 	notsc_setup(NULL);
149 	return 1;
150 }
151 __setup("apicpmtimer", setup_apicpmtimer);
152 #endif
153 
154 static unsigned long mp_lapic_addr __ro_after_init;
155 bool apic_is_disabled __ro_after_init;
156 /* Disable local APIC timer from the kernel commandline or via dmi quirk */
157 static int disable_apic_timer __initdata;
158 /* Local APIC timer works in C2 */
159 int local_apic_timer_c2_ok __ro_after_init;
160 EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok);
161 
162 /*
163  * Debug level, exported for io_apic.c
164  */
165 int apic_verbosity __ro_after_init;
166 
167 int pic_mode __ro_after_init;
168 
169 /* Have we found an MP table */
170 int smp_found_config __ro_after_init;
171 
172 static struct resource lapic_resource = {
173 	.name = "Local APIC",
174 	.flags = IORESOURCE_MEM | IORESOURCE_BUSY,
175 };
176 
177 /* Measured in ticks per HZ. */
178 unsigned int lapic_timer_period = 0;
179 
180 static void apic_pm_activate(void);
181 
182 /*
183  * Get the LAPIC version
184  */
185 static inline int lapic_get_version(void)
186 {
187 	return GET_APIC_VERSION(apic_read(APIC_LVR));
188 }
189 
190 /*
191  * Check, if the APIC is integrated or a separate chip
192  */
193 static inline int lapic_is_integrated(void)
194 {
195 	return APIC_INTEGRATED(lapic_get_version());
196 }
197 
198 /*
199  * Check, whether this is a modern or a first generation APIC
200  */
201 static int modern_apic(void)
202 {
203 	/* AMD systems use old APIC versions, so check the CPU */
204 	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
205 	    boot_cpu_data.x86 >= 0xf)
206 		return 1;
207 
208 	/* Hygon systems use modern APIC */
209 	if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
210 		return 1;
211 
212 	return lapic_get_version() >= 0x14;
213 }
214 
215 /*
216  * right after this call apic become NOOP driven
217  * so apic->write/read doesn't do anything
218  */
219 static void __init apic_disable(void)
220 {
221 	apic_install_driver(&apic_noop);
222 }
223 
224 void native_apic_icr_write(u32 low, u32 id)
225 {
226 	unsigned long flags;
227 
228 	local_irq_save(flags);
229 	apic_write(APIC_ICR2, SET_XAPIC_DEST_FIELD(id));
230 	apic_write(APIC_ICR, low);
231 	local_irq_restore(flags);
232 }
233 
234 u64 native_apic_icr_read(void)
235 {
236 	u32 icr1, icr2;
237 
238 	icr2 = apic_read(APIC_ICR2);
239 	icr1 = apic_read(APIC_ICR);
240 
241 	return icr1 | ((u64)icr2 << 32);
242 }
243 
244 /**
245  * lapic_get_maxlvt - get the maximum number of local vector table entries
246  */
247 int lapic_get_maxlvt(void)
248 {
249 	/*
250 	 * - we always have APIC integrated on 64bit mode
251 	 * - 82489DXs do not report # of LVT entries
252 	 */
253 	return lapic_is_integrated() ? GET_APIC_MAXLVT(apic_read(APIC_LVR)) : 2;
254 }
255 
256 /*
257  * Local APIC timer
258  */
259 
260 /* Clock divisor */
261 #define APIC_DIVISOR 16
262 #define TSC_DIVISOR  8
263 
264 /* i82489DX specific */
265 #define		I82489DX_BASE_DIVIDER		(((0x2) << 18))
266 
267 /*
268  * This function sets up the local APIC timer, with a timeout of
269  * 'clocks' APIC bus clock. During calibration we actually call
270  * this function twice on the boot CPU, once with a bogus timeout
271  * value, second time for real. The other (noncalibrating) CPUs
272  * call this function only once, with the real, calibrated value.
273  *
274  * We do reads before writes even if unnecessary, to get around the
275  * P5 APIC double write bug.
276  */
277 static void __setup_APIC_LVTT(unsigned int clocks, int oneshot, int irqen)
278 {
279 	unsigned int lvtt_value, tmp_value;
280 
281 	lvtt_value = LOCAL_TIMER_VECTOR;
282 	if (!oneshot)
283 		lvtt_value |= APIC_LVT_TIMER_PERIODIC;
284 	else if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
285 		lvtt_value |= APIC_LVT_TIMER_TSCDEADLINE;
286 
287 	/*
288 	 * The i82489DX APIC uses bit 18 and 19 for the base divider.  This
289 	 * overlaps with bit 18 on integrated APICs, but is not documented
290 	 * in the SDM. No problem though. i82489DX equipped systems do not
291 	 * have TSC deadline timer.
292 	 */
293 	if (!lapic_is_integrated())
294 		lvtt_value |= I82489DX_BASE_DIVIDER;
295 
296 	if (!irqen)
297 		lvtt_value |= APIC_LVT_MASKED;
298 
299 	apic_write(APIC_LVTT, lvtt_value);
300 
301 	if (lvtt_value & APIC_LVT_TIMER_TSCDEADLINE) {
302 		/*
303 		 * See Intel SDM: TSC-Deadline Mode chapter. In xAPIC mode,
304 		 * writing to the APIC LVTT and TSC_DEADLINE MSR isn't serialized.
305 		 * According to Intel, MFENCE can do the serialization here.
306 		 */
307 		asm volatile("mfence" : : : "memory");
308 		return;
309 	}
310 
311 	/*
312 	 * Divide PICLK by 16
313 	 */
314 	tmp_value = apic_read(APIC_TDCR);
315 	apic_write(APIC_TDCR,
316 		(tmp_value & ~(APIC_TDR_DIV_1 | APIC_TDR_DIV_TMBASE)) |
317 		APIC_TDR_DIV_16);
318 
319 	if (!oneshot)
320 		apic_write(APIC_TMICT, clocks / APIC_DIVISOR);
321 }
322 
323 /*
324  * Setup extended LVT, AMD specific
325  *
326  * Software should use the LVT offsets the BIOS provides.  The offsets
327  * are determined by the subsystems using it like those for MCE
328  * threshold or IBS.  On K8 only offset 0 (APIC500) and MCE interrupts
329  * are supported. Beginning with family 10h at least 4 offsets are
330  * available.
331  *
332  * Since the offsets must be consistent for all cores, we keep track
333  * of the LVT offsets in software and reserve the offset for the same
334  * vector also to be used on other cores. An offset is freed by
335  * setting the entry to APIC_EILVT_MASKED.
336  *
337  * If the BIOS is right, there should be no conflicts. Otherwise a
338  * "[Firmware Bug]: ..." error message is generated. However, if
339  * software does not properly determines the offsets, it is not
340  * necessarily a BIOS bug.
341  */
342 
343 static atomic_t eilvt_offsets[APIC_EILVT_NR_MAX];
344 
345 static inline int eilvt_entry_is_changeable(unsigned int old, unsigned int new)
346 {
347 	return (old & APIC_EILVT_MASKED)
348 		|| (new == APIC_EILVT_MASKED)
349 		|| ((new & ~APIC_EILVT_MASKED) == old);
350 }
351 
352 static unsigned int reserve_eilvt_offset(int offset, unsigned int new)
353 {
354 	unsigned int rsvd, vector;
355 
356 	if (offset >= APIC_EILVT_NR_MAX)
357 		return ~0;
358 
359 	rsvd = atomic_read(&eilvt_offsets[offset]);
360 	do {
361 		vector = rsvd & ~APIC_EILVT_MASKED;	/* 0: unassigned */
362 		if (vector && !eilvt_entry_is_changeable(vector, new))
363 			/* may not change if vectors are different */
364 			return rsvd;
365 	} while (!atomic_try_cmpxchg(&eilvt_offsets[offset], &rsvd, new));
366 
367 	rsvd = new & ~APIC_EILVT_MASKED;
368 	if (rsvd && rsvd != vector)
369 		pr_info("LVT offset %d assigned for vector 0x%02x\n",
370 			offset, rsvd);
371 
372 	return new;
373 }
374 
375 /*
376  * If mask=1, the LVT entry does not generate interrupts while mask=0
377  * enables the vector. See also the BKDGs. Must be called with
378  * preemption disabled.
379  */
380 
381 int setup_APIC_eilvt(u8 offset, u8 vector, u8 msg_type, u8 mask)
382 {
383 	unsigned long reg = APIC_EILVTn(offset);
384 	unsigned int new, old, reserved;
385 
386 	new = (mask << 16) | (msg_type << 8) | vector;
387 	old = apic_read(reg);
388 	reserved = reserve_eilvt_offset(offset, new);
389 
390 	if (reserved != new) {
391 		pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
392 		       "vector 0x%x, but the register is already in use for "
393 		       "vector 0x%x on another cpu\n",
394 		       smp_processor_id(), reg, offset, new, reserved);
395 		return -EINVAL;
396 	}
397 
398 	if (!eilvt_entry_is_changeable(old, new)) {
399 		pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
400 		       "vector 0x%x, but the register is already in use for "
401 		       "vector 0x%x on this cpu\n",
402 		       smp_processor_id(), reg, offset, new, old);
403 		return -EBUSY;
404 	}
405 
406 	apic_write(reg, new);
407 
408 	return 0;
409 }
410 EXPORT_SYMBOL_GPL(setup_APIC_eilvt);
411 
412 /*
413  * Program the next event, relative to now
414  */
415 static int lapic_next_event(unsigned long delta,
416 			    struct clock_event_device *evt)
417 {
418 	apic_write(APIC_TMICT, delta);
419 	return 0;
420 }
421 
422 static int lapic_next_deadline(unsigned long delta,
423 			       struct clock_event_device *evt)
424 {
425 	u64 tsc;
426 
427 	/* This MSR is special and need a special fence: */
428 	weak_wrmsr_fence();
429 
430 	tsc = rdtsc();
431 	wrmsrq(MSR_IA32_TSC_DEADLINE, tsc + (((u64) delta) * TSC_DIVISOR));
432 	return 0;
433 }
434 
435 static int lapic_timer_shutdown(struct clock_event_device *evt)
436 {
437 	unsigned int v;
438 
439 	/* Lapic used as dummy for broadcast ? */
440 	if (evt->features & CLOCK_EVT_FEAT_DUMMY)
441 		return 0;
442 
443 	v = apic_read(APIC_LVTT);
444 	v |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
445 	apic_write(APIC_LVTT, v);
446 
447 	/*
448 	 * Setting APIC_LVT_MASKED (above) should be enough to tell
449 	 * the hardware that this timer will never fire. But AMD
450 	 * erratum 411 and some Intel CPU behavior circa 2024 say
451 	 * otherwise.  Time for belt and suspenders programming: mask
452 	 * the timer _and_ zero the counter registers:
453 	 */
454 	if (v & APIC_LVT_TIMER_TSCDEADLINE)
455 		wrmsrq(MSR_IA32_TSC_DEADLINE, 0);
456 	else
457 		apic_write(APIC_TMICT, 0);
458 
459 	return 0;
460 }
461 
462 static inline int
463 lapic_timer_set_periodic_oneshot(struct clock_event_device *evt, bool oneshot)
464 {
465 	/* Lapic used as dummy for broadcast ? */
466 	if (evt->features & CLOCK_EVT_FEAT_DUMMY)
467 		return 0;
468 
469 	__setup_APIC_LVTT(lapic_timer_period, oneshot, 1);
470 	return 0;
471 }
472 
473 static int lapic_timer_set_periodic(struct clock_event_device *evt)
474 {
475 	return lapic_timer_set_periodic_oneshot(evt, false);
476 }
477 
478 static int lapic_timer_set_oneshot(struct clock_event_device *evt)
479 {
480 	return lapic_timer_set_periodic_oneshot(evt, true);
481 }
482 
483 /*
484  * Local APIC timer broadcast function
485  */
486 static void lapic_timer_broadcast(const struct cpumask *mask)
487 {
488 #ifdef CONFIG_SMP
489 	__apic_send_IPI_mask(mask, LOCAL_TIMER_VECTOR);
490 #endif
491 }
492 
493 
494 /*
495  * The local apic timer can be used for any function which is CPU local.
496  */
497 static struct clock_event_device lapic_clockevent = {
498 	.name				= "lapic",
499 	.features			= CLOCK_EVT_FEAT_PERIODIC |
500 					  CLOCK_EVT_FEAT_ONESHOT | CLOCK_EVT_FEAT_C3STOP
501 					  | CLOCK_EVT_FEAT_DUMMY,
502 	.shift				= 32,
503 	.set_state_shutdown		= lapic_timer_shutdown,
504 	.set_state_periodic		= lapic_timer_set_periodic,
505 	.set_state_oneshot		= lapic_timer_set_oneshot,
506 	.set_state_oneshot_stopped	= lapic_timer_shutdown,
507 	.set_next_event			= lapic_next_event,
508 	.broadcast			= lapic_timer_broadcast,
509 	.rating				= 100,
510 	.irq				= -1,
511 };
512 static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
513 
514 static const struct x86_cpu_id deadline_match[] __initconst = {
515 	X86_MATCH_VFM_STEPS(INTEL_HASWELL_X,   0x2, 0x2, 0x3a), /* EP */
516 	X86_MATCH_VFM_STEPS(INTEL_HASWELL_X,   0x4, 0x4, 0x0f), /* EX */
517 
518 	X86_MATCH_VFM(INTEL_BROADWELL_X,	0x0b000020),
519 
520 	X86_MATCH_VFM_STEPS(INTEL_BROADWELL_D, 0x2, 0x2, 0x00000011),
521 	X86_MATCH_VFM_STEPS(INTEL_BROADWELL_D, 0x3, 0x3, 0x0700000e),
522 	X86_MATCH_VFM_STEPS(INTEL_BROADWELL_D, 0x4, 0x4, 0x0f00000c),
523 	X86_MATCH_VFM_STEPS(INTEL_BROADWELL_D, 0x5, 0x5, 0x0e000003),
524 
525 	X86_MATCH_VFM_STEPS(INTEL_SKYLAKE_X,   0x3, 0x3, 0x01000136),
526 	X86_MATCH_VFM_STEPS(INTEL_SKYLAKE_X,   0x4, 0x4, 0x02000014),
527 	X86_MATCH_VFM_STEPS(INTEL_SKYLAKE_X,   0x5, 0xf, 0),
528 
529 	X86_MATCH_VFM(INTEL_HASWELL,		0x22),
530 	X86_MATCH_VFM(INTEL_HASWELL_L,		0x20),
531 	X86_MATCH_VFM(INTEL_HASWELL_G,		0x17),
532 
533 	X86_MATCH_VFM(INTEL_BROADWELL,		0x25),
534 	X86_MATCH_VFM(INTEL_BROADWELL_G,	0x17),
535 
536 	X86_MATCH_VFM(INTEL_SKYLAKE_L,		0xb2),
537 	X86_MATCH_VFM(INTEL_SKYLAKE,		0xb2),
538 
539 	X86_MATCH_VFM(INTEL_KABYLAKE_L,		0x52),
540 	X86_MATCH_VFM(INTEL_KABYLAKE,		0x52),
541 
542 	{},
543 };
544 
545 static __init bool apic_validate_deadline_timer(void)
546 {
547 	const struct x86_cpu_id *m;
548 	u32 rev;
549 
550 	if (!boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
551 		return false;
552 	if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
553 		return true;
554 
555 	m = x86_match_cpu(deadline_match);
556 	if (!m)
557 		return true;
558 
559 	rev = (u32)m->driver_data;
560 
561 	if (boot_cpu_data.microcode >= rev)
562 		return true;
563 
564 	setup_clear_cpu_cap(X86_FEATURE_TSC_DEADLINE_TIMER);
565 	pr_err(FW_BUG "TSC_DEADLINE disabled due to Errata; "
566 	       "please update microcode to version: 0x%x (or later)\n", rev);
567 	return false;
568 }
569 
570 /*
571  * Setup the local APIC timer for this CPU. Copy the initialized values
572  * of the boot CPU and register the clock event in the framework.
573  */
574 static void setup_APIC_timer(void)
575 {
576 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
577 
578 	if (this_cpu_has(X86_FEATURE_ARAT)) {
579 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_C3STOP;
580 		/* Make LAPIC timer preferable over percpu HPET */
581 		lapic_clockevent.rating = 150;
582 	}
583 
584 	memcpy(levt, &lapic_clockevent, sizeof(*levt));
585 	levt->cpumask = cpumask_of(smp_processor_id());
586 
587 	if (this_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER)) {
588 		levt->name = "lapic-deadline";
589 		levt->features &= ~(CLOCK_EVT_FEAT_PERIODIC |
590 				    CLOCK_EVT_FEAT_DUMMY);
591 		levt->set_next_event = lapic_next_deadline;
592 		clockevents_config_and_register(levt,
593 						tsc_khz * (1000 / TSC_DIVISOR),
594 						0xF, ~0UL);
595 	} else
596 		clockevents_register_device(levt);
597 
598 	apic_update_vector(smp_processor_id(), LOCAL_TIMER_VECTOR, true);
599 }
600 
601 /*
602  * Install the updated TSC frequency from recalibration at the TSC
603  * deadline clockevent devices.
604  */
605 static void __lapic_update_tsc_freq(void *info)
606 {
607 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
608 
609 	if (!this_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
610 		return;
611 
612 	clockevents_update_freq(levt, tsc_khz * (1000 / TSC_DIVISOR));
613 }
614 
615 void lapic_update_tsc_freq(void)
616 {
617 	/*
618 	 * The clockevent device's ->mult and ->shift can both be
619 	 * changed. In order to avoid races, schedule the frequency
620 	 * update code on each CPU.
621 	 */
622 	on_each_cpu(__lapic_update_tsc_freq, NULL, 0);
623 }
624 
625 /*
626  * In this functions we calibrate APIC bus clocks to the external timer.
627  *
628  * We want to do the calibration only once since we want to have local timer
629  * irqs synchronous. CPUs connected by the same APIC bus have the very same bus
630  * frequency.
631  *
632  * This was previously done by reading the PIT/HPET and waiting for a wrap
633  * around to find out, that a tick has elapsed. I have a box, where the PIT
634  * readout is broken, so it never gets out of the wait loop again. This was
635  * also reported by others.
636  *
637  * Monitoring the jiffies value is inaccurate and the clockevents
638  * infrastructure allows us to do a simple substitution of the interrupt
639  * handler.
640  *
641  * The calibration routine also uses the pm_timer when possible, as the PIT
642  * happens to run way too slow (factor 2.3 on my VAIO CoreDuo, which goes
643  * back to normal later in the boot process).
644  */
645 
646 #define LAPIC_CAL_LOOPS		(HZ/10)
647 
648 static __initdata int lapic_cal_loops = -1;
649 static __initdata long lapic_cal_t1, lapic_cal_t2;
650 static __initdata unsigned long long lapic_cal_tsc1, lapic_cal_tsc2;
651 static __initdata u32 lapic_cal_pm1, lapic_cal_pm2;
652 static __initdata unsigned long lapic_cal_j1, lapic_cal_j2;
653 
654 /*
655  * Temporary interrupt handler and polled calibration function.
656  */
657 static void __init lapic_cal_handler(struct clock_event_device *dev)
658 {
659 	unsigned long long tsc = 0;
660 	long tapic = apic_read(APIC_TMCCT);
661 	u32 pm = acpi_pm_read_early();
662 
663 	if (boot_cpu_has(X86_FEATURE_TSC))
664 		tsc = rdtsc();
665 
666 	switch (lapic_cal_loops++) {
667 	case 0:
668 		lapic_cal_t1 = tapic;
669 		lapic_cal_tsc1 = tsc;
670 		lapic_cal_pm1 = pm;
671 		lapic_cal_j1 = jiffies;
672 		break;
673 
674 	case LAPIC_CAL_LOOPS:
675 		lapic_cal_t2 = tapic;
676 		lapic_cal_tsc2 = tsc;
677 		if (pm < lapic_cal_pm1)
678 			pm += ACPI_PM_OVRRUN;
679 		lapic_cal_pm2 = pm;
680 		lapic_cal_j2 = jiffies;
681 		break;
682 	}
683 }
684 
685 static int __init
686 calibrate_by_pmtimer(u32 deltapm, long *delta, long *deltatsc)
687 {
688 	const long pm_100ms = PMTMR_TICKS_PER_SEC / 10;
689 	const long pm_thresh = pm_100ms / 100;
690 	unsigned long mult;
691 	u64 res;
692 
693 #ifndef CONFIG_X86_PM_TIMER
694 	return -1;
695 #endif
696 
697 	apic_pr_verbose("... PM-Timer delta = %u\n", deltapm);
698 
699 	/* Check, if the PM timer is available */
700 	if (!deltapm)
701 		return -1;
702 
703 	mult = clocksource_hz2mult(PMTMR_TICKS_PER_SEC, 22);
704 
705 	if (deltapm > (pm_100ms - pm_thresh) &&
706 	    deltapm < (pm_100ms + pm_thresh)) {
707 		apic_pr_verbose("... PM-Timer result ok\n");
708 		return 0;
709 	}
710 
711 	res = (((u64)deltapm) *  mult) >> 22;
712 	do_div(res, 1000000);
713 	pr_warn("APIC calibration not consistent with PM-Timer: %ldms instead of 100ms\n",
714 		(long)res);
715 
716 	/* Correct the lapic counter value */
717 	res = (((u64)(*delta)) * pm_100ms);
718 	do_div(res, deltapm);
719 	pr_info("APIC delta adjusted to PM-Timer: "
720 		"%lu (%ld)\n", (unsigned long)res, *delta);
721 	*delta = (long)res;
722 
723 	/* Correct the tsc counter value */
724 	if (boot_cpu_has(X86_FEATURE_TSC)) {
725 		res = (((u64)(*deltatsc)) * pm_100ms);
726 		do_div(res, deltapm);
727 		apic_pr_verbose("TSC delta adjusted to PM-Timer: %lu (%ld)\n",
728 				(unsigned long)res, *deltatsc);
729 		*deltatsc = (long)res;
730 	}
731 
732 	return 0;
733 }
734 
735 static int __init lapic_init_clockevent(void)
736 {
737 	if (!lapic_timer_period)
738 		return -1;
739 
740 	/* Calculate the scaled math multiplication factor */
741 	lapic_clockevent.mult = div_sc(lapic_timer_period/APIC_DIVISOR,
742 					TICK_NSEC, lapic_clockevent.shift);
743 	lapic_clockevent.max_delta_ns =
744 		clockevent_delta2ns(0x7FFFFFFF, &lapic_clockevent);
745 	lapic_clockevent.max_delta_ticks = 0x7FFFFFFF;
746 	lapic_clockevent.min_delta_ns =
747 		clockevent_delta2ns(0xF, &lapic_clockevent);
748 	lapic_clockevent.min_delta_ticks = 0xF;
749 
750 	return 0;
751 }
752 
753 bool __init apic_needs_pit(void)
754 {
755 	/*
756 	 * If the frequencies are not known, PIT is required for both TSC
757 	 * and apic timer calibration.
758 	 */
759 	if (!tsc_khz || !cpu_khz)
760 		return true;
761 
762 	/* Is there an APIC at all or is it disabled? */
763 	if (!boot_cpu_has(X86_FEATURE_APIC) || apic_is_disabled)
764 		return true;
765 
766 	/*
767 	 * If interrupt delivery mode is legacy PIC or virtual wire without
768 	 * configuration, the local APIC timer won't be set up. Make sure
769 	 * that the PIT is initialized.
770 	 */
771 	if (apic_intr_mode == APIC_PIC ||
772 	    apic_intr_mode == APIC_VIRTUAL_WIRE_NO_CONFIG)
773 		return true;
774 
775 	/* Virt guests may lack ARAT, but still have DEADLINE */
776 	if (!boot_cpu_has(X86_FEATURE_ARAT))
777 		return true;
778 
779 	/* Deadline timer is based on TSC so no further PIT action required */
780 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
781 		return false;
782 
783 	/* APIC timer disabled? */
784 	if (disable_apic_timer)
785 		return true;
786 	/*
787 	 * The APIC timer frequency is known already, no PIT calibration
788 	 * required. If unknown, let the PIT be initialized.
789 	 */
790 	return lapic_timer_period == 0;
791 }
792 
793 static int __init calibrate_APIC_clock(void)
794 {
795 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
796 	u64 tsc_perj = 0, tsc_start = 0;
797 	long delta_tsc_khz, bus_khz;
798 	unsigned long jif_start;
799 	unsigned long deltaj;
800 	long delta, deltatsc;
801 	int pm_referenced = 0;
802 
803 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
804 		return 0;
805 
806 	/*
807 	 * Check if lapic timer has already been calibrated by platform
808 	 * specific routine, such as tsc calibration code. If so just fill
809 	 * in the clockevent structure and return.
810 	 */
811 	if (!lapic_init_clockevent()) {
812 		apic_pr_verbose("lapic timer already calibrated %d\n", lapic_timer_period);
813 		/*
814 		 * Direct calibration methods must have an always running
815 		 * local APIC timer, no need for broadcast timer.
816 		 */
817 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
818 		return 0;
819 	}
820 
821 	apic_pr_verbose("Using local APIC timer interrupts. Calibrating APIC timer ...\n");
822 
823 	/*
824 	 * There are platforms w/o global clockevent devices. Instead of
825 	 * making the calibration conditional on that, use a polling based
826 	 * approach everywhere.
827 	 */
828 	local_irq_disable();
829 
830 	/*
831 	 * Setup the APIC counter to maximum. There is no way the lapic
832 	 * can underflow in the 100ms detection time frame
833 	 */
834 	__setup_APIC_LVTT(0xffffffff, 0, 0);
835 
836 	/*
837 	 * Methods to terminate the calibration loop:
838 	 *  1) Global clockevent if available (jiffies)
839 	 *  2) TSC if available and frequency is known
840 	 */
841 	jif_start = READ_ONCE(jiffies);
842 
843 	if (tsc_khz) {
844 		tsc_start = rdtsc();
845 		tsc_perj = div_u64((u64)tsc_khz * 1000, HZ);
846 	}
847 
848 	/*
849 	 * Enable interrupts so the tick can fire, if a global
850 	 * clockevent device is available
851 	 */
852 	local_irq_enable();
853 
854 	while (lapic_cal_loops <= LAPIC_CAL_LOOPS) {
855 		/* Wait for a tick to elapse */
856 		while (1) {
857 			if (tsc_khz) {
858 				u64 tsc_now = rdtsc();
859 				if ((tsc_now - tsc_start) >= tsc_perj) {
860 					tsc_start += tsc_perj;
861 					break;
862 				}
863 			} else {
864 				unsigned long jif_now = READ_ONCE(jiffies);
865 
866 				if (time_after(jif_now, jif_start)) {
867 					jif_start = jif_now;
868 					break;
869 				}
870 			}
871 			cpu_relax();
872 		}
873 
874 		/* Invoke the calibration routine */
875 		local_irq_disable();
876 		lapic_cal_handler(NULL);
877 		local_irq_enable();
878 	}
879 
880 	local_irq_disable();
881 
882 	/* Build delta t1-t2 as apic timer counts down */
883 	delta = lapic_cal_t1 - lapic_cal_t2;
884 	apic_pr_verbose("... lapic delta = %ld\n", delta);
885 
886 	deltatsc = (long)(lapic_cal_tsc2 - lapic_cal_tsc1);
887 
888 	/* we trust the PM based calibration if possible */
889 	pm_referenced = !calibrate_by_pmtimer(lapic_cal_pm2 - lapic_cal_pm1,
890 					&delta, &deltatsc);
891 
892 	lapic_timer_period = (delta * APIC_DIVISOR) / LAPIC_CAL_LOOPS;
893 	lapic_init_clockevent();
894 
895 	apic_pr_verbose("..... delta %ld\n", delta);
896 	apic_pr_verbose("..... mult: %u\n", lapic_clockevent.mult);
897 	apic_pr_verbose("..... calibration result: %u\n", lapic_timer_period);
898 
899 	if (boot_cpu_has(X86_FEATURE_TSC)) {
900 		delta_tsc_khz = (deltatsc * HZ) / (1000 * LAPIC_CAL_LOOPS);
901 
902 		apic_pr_verbose("..... CPU clock speed is %ld.%03ld MHz.\n",
903 				delta_tsc_khz / 1000, delta_tsc_khz % 1000);
904 	}
905 
906 	bus_khz = (long)lapic_timer_period * HZ / 1000;
907 	apic_pr_verbose("..... host bus clock speed is %ld.%03ld MHz.\n",
908 			bus_khz / 1000, bus_khz % 1000);
909 
910 	/*
911 	 * Do a sanity check on the APIC calibration result
912 	 */
913 	if (lapic_timer_period < (1000000 / HZ)) {
914 		local_irq_enable();
915 		pr_warn("APIC frequency too slow, disabling apic timer\n");
916 		return -1;
917 	}
918 
919 	levt->features &= ~CLOCK_EVT_FEAT_DUMMY;
920 
921 	/*
922 	 * PM timer calibration failed or not turned on so lets try APIC
923 	 * timer based calibration, if a global clockevent device is
924 	 * available.
925 	 */
926 	if (!pm_referenced && global_clock_event) {
927 		apic_pr_verbose("... verify APIC timer\n");
928 
929 		/*
930 		 * Setup the apic timer manually
931 		 */
932 		levt->event_handler = lapic_cal_handler;
933 		lapic_timer_set_periodic(levt);
934 		lapic_cal_loops = -1;
935 
936 		/* Let the interrupts run */
937 		local_irq_enable();
938 
939 		while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
940 			cpu_relax();
941 
942 		/* Stop the lapic timer */
943 		local_irq_disable();
944 		lapic_timer_shutdown(levt);
945 
946 		/* Jiffies delta */
947 		deltaj = lapic_cal_j2 - lapic_cal_j1;
948 		apic_pr_verbose("... jiffies delta = %lu\n", deltaj);
949 
950 		/* Check, if the jiffies result is consistent */
951 		if (deltaj >= LAPIC_CAL_LOOPS-2 && deltaj <= LAPIC_CAL_LOOPS+2)
952 			apic_pr_verbose("... jiffies result ok\n");
953 		else
954 			levt->features |= CLOCK_EVT_FEAT_DUMMY;
955 	}
956 	local_irq_enable();
957 
958 	if (levt->features & CLOCK_EVT_FEAT_DUMMY) {
959 		pr_warn("APIC timer disabled due to verification failure\n");
960 		return -1;
961 	}
962 
963 	return 0;
964 }
965 
966 /*
967  * Setup the boot APIC
968  *
969  * Calibrate and verify the result.
970  */
971 void __init setup_boot_APIC_clock(void)
972 {
973 	/*
974 	 * The local apic timer can be disabled via the kernel
975 	 * commandline or from the CPU detection code. Register the lapic
976 	 * timer as a dummy clock event source on SMP systems, so the
977 	 * broadcast mechanism is used. On UP systems simply ignore it.
978 	 */
979 	if (disable_apic_timer) {
980 		pr_info("Disabling APIC timer\n");
981 		/* No broadcast on UP ! */
982 		if (num_possible_cpus() > 1) {
983 			lapic_clockevent.mult = 1;
984 			setup_APIC_timer();
985 		}
986 		return;
987 	}
988 
989 	if (calibrate_APIC_clock()) {
990 		/* No broadcast on UP ! */
991 		if (num_possible_cpus() > 1)
992 			setup_APIC_timer();
993 		return;
994 	}
995 
996 	/*
997 	 * If nmi_watchdog is set to IO_APIC, we need the
998 	 * PIT/HPET going.  Otherwise register lapic as a dummy
999 	 * device.
1000 	 */
1001 	lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
1002 
1003 	/* Setup the lapic or request the broadcast */
1004 	setup_APIC_timer();
1005 	amd_e400_c1e_apic_setup();
1006 }
1007 
1008 void setup_secondary_APIC_clock(void)
1009 {
1010 	setup_APIC_timer();
1011 	amd_e400_c1e_apic_setup();
1012 }
1013 
1014 /*
1015  * The guts of the apic timer interrupt
1016  */
1017 static void local_apic_timer_interrupt(void)
1018 {
1019 	struct clock_event_device *evt = this_cpu_ptr(&lapic_events);
1020 
1021 	/*
1022 	 * Normally we should not be here till LAPIC has been initialized but
1023 	 * in some cases like kdump, its possible that there is a pending LAPIC
1024 	 * timer interrupt from previous kernel's context and is delivered in
1025 	 * new kernel the moment interrupts are enabled.
1026 	 *
1027 	 * Interrupts are enabled early and LAPIC is setup much later, hence
1028 	 * its possible that when we get here evt->event_handler is NULL.
1029 	 * Check for event_handler being NULL and discard the interrupt as
1030 	 * spurious.
1031 	 */
1032 	if (!evt->event_handler) {
1033 		pr_warn("Spurious LAPIC timer interrupt on cpu %d\n",
1034 			smp_processor_id());
1035 		/* Switch it off */
1036 		lapic_timer_shutdown(evt);
1037 		return;
1038 	}
1039 
1040 	/*
1041 	 * the NMI deadlock-detector uses this.
1042 	 */
1043 	inc_irq_stat(apic_timer_irqs);
1044 
1045 	evt->event_handler(evt);
1046 }
1047 
1048 /*
1049  * Local APIC timer interrupt. This is the most natural way for doing
1050  * local interrupts, but local timer interrupts can be emulated by
1051  * broadcast interrupts too. [in case the hw doesn't support APIC timers]
1052  *
1053  * [ if a single-CPU system runs an SMP kernel then we call the local
1054  *   interrupt as well. Thus we cannot inline the local irq ... ]
1055  */
1056 DEFINE_IDTENTRY_SYSVEC(sysvec_apic_timer_interrupt)
1057 {
1058 	struct pt_regs *old_regs = set_irq_regs(regs);
1059 
1060 	apic_eoi();
1061 	trace_local_timer_entry(LOCAL_TIMER_VECTOR);
1062 	local_apic_timer_interrupt();
1063 	trace_local_timer_exit(LOCAL_TIMER_VECTOR);
1064 
1065 	set_irq_regs(old_regs);
1066 }
1067 
1068 /*
1069  * Local APIC start and shutdown
1070  */
1071 
1072 /**
1073  * clear_local_APIC - shutdown the local APIC
1074  *
1075  * This is called, when a CPU is disabled and before rebooting, so the state of
1076  * the local APIC has no dangling leftovers. Also used to cleanout any BIOS
1077  * leftovers during boot.
1078  */
1079 void clear_local_APIC(void)
1080 {
1081 	int maxlvt;
1082 	u32 v;
1083 
1084 	if (!apic_accessible())
1085 		return;
1086 
1087 	maxlvt = lapic_get_maxlvt();
1088 	/*
1089 	 * Masking an LVT entry can trigger a local APIC error
1090 	 * if the vector is zero. Mask LVTERR first to prevent this.
1091 	 */
1092 	if (maxlvt >= 3) {
1093 		v = ERROR_APIC_VECTOR; /* any non-zero vector will do */
1094 		apic_write(APIC_LVTERR, v | APIC_LVT_MASKED);
1095 	}
1096 	/*
1097 	 * Careful: we have to set masks only first to deassert
1098 	 * any level-triggered sources.
1099 	 */
1100 	v = apic_read(APIC_LVTT);
1101 	apic_write(APIC_LVTT, v | APIC_LVT_MASKED);
1102 	v = apic_read(APIC_LVT0);
1103 	apic_write(APIC_LVT0, v | APIC_LVT_MASKED);
1104 	v = apic_read(APIC_LVT1);
1105 	apic_write(APIC_LVT1, v | APIC_LVT_MASKED);
1106 	if (maxlvt >= 4) {
1107 		v = apic_read(APIC_LVTPC);
1108 		apic_write(APIC_LVTPC, v | APIC_LVT_MASKED);
1109 	}
1110 
1111 	/* lets not touch this if we didn't frob it */
1112 #ifdef CONFIG_X86_THERMAL_VECTOR
1113 	if (maxlvt >= 5) {
1114 		v = apic_read(APIC_LVTTHMR);
1115 		apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED);
1116 	}
1117 #endif
1118 #ifdef CONFIG_X86_MCE_INTEL
1119 	if (maxlvt >= 6) {
1120 		v = apic_read(APIC_LVTCMCI);
1121 		if (!(v & APIC_LVT_MASKED))
1122 			apic_write(APIC_LVTCMCI, v | APIC_LVT_MASKED);
1123 	}
1124 #endif
1125 
1126 	/*
1127 	 * Clean APIC state for other OSs:
1128 	 */
1129 	apic_write(APIC_LVTT, APIC_LVT_MASKED);
1130 	apic_write(APIC_LVT0, APIC_LVT_MASKED);
1131 	apic_write(APIC_LVT1, APIC_LVT_MASKED);
1132 	if (maxlvt >= 3)
1133 		apic_write(APIC_LVTERR, APIC_LVT_MASKED);
1134 	if (maxlvt >= 4)
1135 		apic_write(APIC_LVTPC, APIC_LVT_MASKED);
1136 
1137 	/* Integrated APIC (!82489DX) ? */
1138 	if (lapic_is_integrated()) {
1139 		if (maxlvt > 3)
1140 			/* Clear ESR due to Pentium errata 3AP and 11AP */
1141 			apic_write(APIC_ESR, 0);
1142 		apic_read(APIC_ESR);
1143 	}
1144 }
1145 
1146 /**
1147  * apic_soft_disable - Clears and software disables the local APIC on hotplug
1148  *
1149  * Contrary to disable_local_APIC() this does not touch the enable bit in
1150  * MSR_IA32_APICBASE. Clearing that bit on systems based on the 3 wire APIC
1151  * bus would require a hardware reset as the APIC would lose track of bus
1152  * arbitration. On systems with FSB delivery APICBASE could be disabled,
1153  * but it has to be guaranteed that no interrupt is sent to the APIC while
1154  * in that state and it's not clear from the SDM whether it still responds
1155  * to INIT/SIPI messages. Stay on the safe side and use software disable.
1156  */
1157 void apic_soft_disable(void)
1158 {
1159 	u32 value;
1160 
1161 	clear_local_APIC();
1162 
1163 	/* Soft disable APIC (implies clearing of registers for 82489DX!). */
1164 	value = apic_read(APIC_SPIV);
1165 	value &= ~APIC_SPIV_APIC_ENABLED;
1166 	apic_write(APIC_SPIV, value);
1167 }
1168 
1169 /**
1170  * disable_local_APIC - clear and disable the local APIC
1171  */
1172 void disable_local_APIC(void)
1173 {
1174 	if (!apic_accessible())
1175 		return;
1176 
1177 	if (apic->teardown)
1178 		apic->teardown();
1179 
1180 	apic_soft_disable();
1181 
1182 #ifdef CONFIG_X86_32
1183 	/*
1184 	 * When LAPIC was disabled by the BIOS and enabled by the kernel,
1185 	 * restore the disabled state.
1186 	 */
1187 	if (enabled_via_apicbase) {
1188 		unsigned int l, h;
1189 
1190 		rdmsr(MSR_IA32_APICBASE, l, h);
1191 		l &= ~MSR_IA32_APICBASE_ENABLE;
1192 		wrmsr(MSR_IA32_APICBASE, l, h);
1193 	}
1194 #endif
1195 }
1196 
1197 /*
1198  * If Linux enabled the LAPIC against the BIOS default disable it down before
1199  * re-entering the BIOS on shutdown.  Otherwise the BIOS may get confused and
1200  * not power-off.  Additionally clear all LVT entries before disable_local_APIC
1201  * for the case where Linux didn't enable the LAPIC.
1202  */
1203 void lapic_shutdown(void)
1204 {
1205 	unsigned long flags;
1206 
1207 	if (!boot_cpu_has(X86_FEATURE_APIC) && !apic_from_smp_config())
1208 		return;
1209 
1210 	local_irq_save(flags);
1211 
1212 #ifdef CONFIG_X86_32
1213 	if (!enabled_via_apicbase)
1214 		clear_local_APIC();
1215 	else
1216 #endif
1217 		disable_local_APIC();
1218 
1219 
1220 	local_irq_restore(flags);
1221 }
1222 
1223 /**
1224  * sync_Arb_IDs - synchronize APIC bus arbitration IDs
1225  */
1226 void __init sync_Arb_IDs(void)
1227 {
1228 	/*
1229 	 * Unsupported on P4 - see Intel Dev. Manual Vol. 3, Ch. 8.6.1 And not
1230 	 * needed on AMD.
1231 	 */
1232 	if (modern_apic() || boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
1233 		return;
1234 
1235 	/*
1236 	 * Wait for idle.
1237 	 */
1238 	apic_wait_icr_idle();
1239 
1240 	apic_pr_debug("Synchronizing Arb IDs.\n");
1241 	apic_write(APIC_ICR, APIC_DEST_ALLINC | APIC_INT_LEVELTRIG | APIC_DM_INIT);
1242 }
1243 
1244 enum apic_intr_mode_id apic_intr_mode __ro_after_init;
1245 
1246 static int __init __apic_intr_mode_select(void)
1247 {
1248 	/* Check kernel option */
1249 	if (apic_is_disabled) {
1250 		pr_info("APIC disabled via kernel command line\n");
1251 		return APIC_PIC;
1252 	}
1253 
1254 	/* Check BIOS */
1255 #ifdef CONFIG_X86_64
1256 	/* On 64-bit, the APIC must be integrated, Check local APIC only */
1257 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1258 		apic_is_disabled = true;
1259 		pr_info("APIC disabled by BIOS\n");
1260 		return APIC_PIC;
1261 	}
1262 #else
1263 	/* On 32-bit, the APIC may be integrated APIC or 82489DX */
1264 
1265 	/* Neither 82489DX nor integrated APIC ? */
1266 	if (!boot_cpu_has(X86_FEATURE_APIC) && !smp_found_config) {
1267 		apic_is_disabled = true;
1268 		return APIC_PIC;
1269 	}
1270 
1271 	/* If the BIOS pretends there is an integrated APIC ? */
1272 	if (!boot_cpu_has(X86_FEATURE_APIC) &&
1273 		APIC_INTEGRATED(boot_cpu_apic_version)) {
1274 		apic_is_disabled = true;
1275 		pr_err(FW_BUG "Local APIC not detected, force emulation\n");
1276 		return APIC_PIC;
1277 	}
1278 #endif
1279 
1280 	/* Check MP table or ACPI MADT configuration */
1281 	if (!smp_found_config) {
1282 		disable_ioapic_support();
1283 		if (!acpi_lapic) {
1284 			pr_info("APIC: ACPI MADT or MP tables are not detected\n");
1285 			return APIC_VIRTUAL_WIRE_NO_CONFIG;
1286 		}
1287 		return APIC_VIRTUAL_WIRE;
1288 	}
1289 
1290 #ifdef CONFIG_SMP
1291 	/* If SMP should be disabled, then really disable it! */
1292 	if (!setup_max_cpus) {
1293 		pr_info("APIC: SMP mode deactivated\n");
1294 		return APIC_SYMMETRIC_IO_NO_ROUTING;
1295 	}
1296 #endif
1297 
1298 	return APIC_SYMMETRIC_IO;
1299 }
1300 
1301 /* Select the interrupt delivery mode for the BSP */
1302 void __init apic_intr_mode_select(void)
1303 {
1304 	apic_intr_mode = __apic_intr_mode_select();
1305 }
1306 
1307 /*
1308  * An initial setup of the virtual wire mode.
1309  */
1310 void __init init_bsp_APIC(void)
1311 {
1312 	unsigned int value;
1313 
1314 	/*
1315 	 * Don't do the setup now if we have a SMP BIOS as the
1316 	 * through-I/O-APIC virtual wire mode might be active.
1317 	 */
1318 	if (smp_found_config || !boot_cpu_has(X86_FEATURE_APIC))
1319 		return;
1320 
1321 	/*
1322 	 * Do not trust the local APIC being empty at bootup.
1323 	 */
1324 	clear_local_APIC();
1325 
1326 	/*
1327 	 * Enable APIC.
1328 	 */
1329 	value = apic_read(APIC_SPIV);
1330 	value &= ~APIC_VECTOR_MASK;
1331 	value |= APIC_SPIV_APIC_ENABLED;
1332 
1333 #ifdef CONFIG_X86_32
1334 	/* This bit is reserved on P4/Xeon and should be cleared */
1335 	if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
1336 	    (boot_cpu_data.x86 == 15))
1337 		value &= ~APIC_SPIV_FOCUS_DISABLED;
1338 	else
1339 #endif
1340 		value |= APIC_SPIV_FOCUS_DISABLED;
1341 	value |= SPURIOUS_APIC_VECTOR;
1342 	apic_write(APIC_SPIV, value);
1343 
1344 	/*
1345 	 * Set up the virtual wire mode.
1346 	 */
1347 	apic_write(APIC_LVT0, APIC_DM_EXTINT);
1348 	value = APIC_DM_NMI;
1349 	if (!lapic_is_integrated())		/* 82489DX */
1350 		value |= APIC_LVT_LEVEL_TRIGGER;
1351 	if (apic_extnmi == APIC_EXTNMI_NONE)
1352 		value |= APIC_LVT_MASKED;
1353 	apic_write(APIC_LVT1, value);
1354 }
1355 
1356 static void __init apic_bsp_setup(bool upmode);
1357 
1358 /* Init the interrupt delivery mode for the BSP */
1359 void __init apic_intr_mode_init(void)
1360 {
1361 	bool upmode = IS_ENABLED(CONFIG_UP_LATE_INIT);
1362 
1363 	switch (apic_intr_mode) {
1364 	case APIC_PIC:
1365 		pr_info("APIC: Keep in PIC mode(8259)\n");
1366 		return;
1367 	case APIC_VIRTUAL_WIRE:
1368 		pr_info("APIC: Switch to virtual wire mode setup\n");
1369 		break;
1370 	case APIC_VIRTUAL_WIRE_NO_CONFIG:
1371 		pr_info("APIC: Switch to virtual wire mode setup with no configuration\n");
1372 		upmode = true;
1373 		break;
1374 	case APIC_SYMMETRIC_IO:
1375 		pr_info("APIC: Switch to symmetric I/O mode setup\n");
1376 		break;
1377 	case APIC_SYMMETRIC_IO_NO_ROUTING:
1378 		pr_info("APIC: Switch to symmetric I/O mode setup in no SMP routine\n");
1379 		break;
1380 	}
1381 
1382 	x86_64_probe_apic();
1383 
1384 	if (x86_platform.apic_post_init)
1385 		x86_platform.apic_post_init();
1386 
1387 	apic_bsp_setup(upmode);
1388 }
1389 
1390 static void lapic_setup_esr(void)
1391 {
1392 	unsigned int oldvalue, value, maxlvt;
1393 
1394 	if (!lapic_is_integrated()) {
1395 		pr_info("No ESR for 82489DX.\n");
1396 		return;
1397 	}
1398 
1399 	if (apic->disable_esr) {
1400 		/*
1401 		 * Something untraceable is creating bad interrupts on
1402 		 * secondary quads ... for the moment, just leave the
1403 		 * ESR disabled - we can't do anything useful with the
1404 		 * errors anyway - mbligh
1405 		 */
1406 		pr_info("Leaving ESR disabled.\n");
1407 		return;
1408 	}
1409 
1410 	maxlvt = lapic_get_maxlvt();
1411 	if (maxlvt > 3)		/* Due to the Pentium erratum 3AP. */
1412 		apic_write(APIC_ESR, 0);
1413 	oldvalue = apic_read(APIC_ESR);
1414 
1415 	/* enables sending errors */
1416 	value = ERROR_APIC_VECTOR;
1417 	apic_write(APIC_LVTERR, value);
1418 
1419 	/*
1420 	 * spec says clear errors after enabling vector.
1421 	 */
1422 	if (maxlvt > 3)
1423 		apic_write(APIC_ESR, 0);
1424 	value = apic_read(APIC_ESR);
1425 	if (value != oldvalue) {
1426 		apic_pr_verbose("ESR value before enabling vector: 0x%08x  after: 0x%08x\n",
1427 				oldvalue, value);
1428 	}
1429 }
1430 
1431 #define APIC_IR_REGS		APIC_ISR_NR
1432 #define APIC_IR_BITS		(APIC_IR_REGS * 32)
1433 #define APIC_IR_MAPSIZE		(APIC_IR_BITS / BITS_PER_LONG)
1434 
1435 union apic_ir {
1436 	unsigned long	map[APIC_IR_MAPSIZE];
1437 	u32		regs[APIC_IR_REGS];
1438 };
1439 
1440 static bool apic_check_and_eoi_isr(union apic_ir *isr)
1441 {
1442 	int i, bit;
1443 
1444 	/* Read the ISRs */
1445 	for (i = 0; i < APIC_IR_REGS; i++)
1446 		isr->regs[i] = apic_read(APIC_ISR + i * 0x10);
1447 
1448 	/* If the ISR map empty, nothing to do here. */
1449 	if (bitmap_empty(isr->map, APIC_IR_BITS))
1450 		return true;
1451 
1452 	/*
1453 	 * There can be multiple ISR bits set when a high priority
1454 	 * interrupt preempted a lower priority one. Issue an EOI for each
1455 	 * set bit. The priority traversal order does not matter as there
1456 	 * can't be new ISR bits raised at this point. What matters is that
1457 	 * an EOI is issued for each ISR bit.
1458 	 */
1459 	for_each_set_bit(bit, isr->map, APIC_IR_BITS)
1460 		apic_eoi();
1461 
1462 	/* Reread the ISRs, they should be empty now */
1463 	for (i = 0; i < APIC_IR_REGS; i++)
1464 		isr->regs[i] = apic_read(APIC_ISR + i * 0x10);
1465 
1466 	return bitmap_empty(isr->map, APIC_IR_BITS);
1467 }
1468 
1469 /*
1470  * If a CPU services an interrupt and crashes before issuing EOI to the
1471  * local APIC, the corresponding ISR bit is still set when the crashing CPU
1472  * jumps into a crash kernel. Read the ISR and issue an EOI for each set
1473  * bit to acknowledge it as otherwise these slots would be locked forever
1474  * waiting for an EOI.
1475  *
1476  * If there are pending bits in the IRR, then they won't be converted into
1477  * ISR bits as the CPU has interrupts disabled. They will be delivered once
1478  * the CPU enables interrupts and there is nothing which can prevent that.
1479  *
1480  * In the worst case this results in spurious interrupt warnings.
1481  */
1482 static void apic_clear_isr(void)
1483 {
1484 	union apic_ir ir;
1485 	unsigned int i;
1486 
1487 	if (!apic_check_and_eoi_isr(&ir))
1488 		pr_warn("APIC: Stale ISR: %256pb\n", ir.map);
1489 
1490 	for (i = 0; i < APIC_IR_REGS; i++)
1491 		ir.regs[i] = apic_read(APIC_IRR + i * 0x10);
1492 
1493 	if (!bitmap_empty(ir.map, APIC_IR_BITS))
1494 		pr_warn("APIC: Stale IRR: %256pb\n", ir.map);
1495 }
1496 
1497 /**
1498  * setup_local_APIC - setup the local APIC
1499  *
1500  * Used to setup local APIC while initializing BSP or bringing up APs.
1501  * Always called with preemption disabled.
1502  */
1503 static void setup_local_APIC(void)
1504 {
1505 	int cpu = smp_processor_id();
1506 	unsigned int value;
1507 
1508 	if (apic_is_disabled) {
1509 		disable_ioapic_support();
1510 		return;
1511 	}
1512 
1513 	if (apic->setup)
1514 		apic->setup();
1515 
1516 	/*
1517 	 * If this comes from kexec/kcrash the APIC might be enabled in
1518 	 * SPIV. Soft disable it before doing further initialization.
1519 	 */
1520 	value = apic_read(APIC_SPIV);
1521 	value &= ~APIC_SPIV_APIC_ENABLED;
1522 	apic_write(APIC_SPIV, value);
1523 
1524 #ifdef CONFIG_X86_32
1525 	/* Pound the ESR really hard over the head with a big hammer - mbligh */
1526 	if (lapic_is_integrated() && apic->disable_esr) {
1527 		apic_write(APIC_ESR, 0);
1528 		apic_write(APIC_ESR, 0);
1529 		apic_write(APIC_ESR, 0);
1530 		apic_write(APIC_ESR, 0);
1531 	}
1532 #endif
1533 	/*
1534 	 * Intel recommends to set DFR, LDR and TPR before enabling
1535 	 * an APIC.  See e.g. "AP-388 82489DX User's Manual" (Intel
1536 	 * document number 292116).
1537 	 *
1538 	 * Except for APICs which operate in physical destination mode.
1539 	 */
1540 	if (apic->init_apic_ldr)
1541 		apic->init_apic_ldr();
1542 
1543 	/*
1544 	 * Set Task Priority to 'accept all except vectors 0-31'.  An APIC
1545 	 * vector in the 16-31 range could be delivered if TPR == 0, but we
1546 	 * would think it's an exception and terrible things will happen.  We
1547 	 * never change this later on.
1548 	 */
1549 	value = apic_read(APIC_TASKPRI);
1550 	value &= ~APIC_TPRI_MASK;
1551 	value |= 0x10;
1552 	apic_write(APIC_TASKPRI, value);
1553 
1554 	apic_clear_isr();
1555 
1556 	/*
1557 	 * Now that we are all set up, enable the APIC
1558 	 */
1559 	value = apic_read(APIC_SPIV);
1560 	value &= ~APIC_VECTOR_MASK;
1561 	/*
1562 	 * Enable APIC
1563 	 */
1564 	value |= APIC_SPIV_APIC_ENABLED;
1565 
1566 #ifdef CONFIG_X86_32
1567 	/*
1568 	 * Some unknown Intel IO/APIC (or APIC) errata is biting us with
1569 	 * certain networking cards. If high frequency interrupts are
1570 	 * happening on a particular IOAPIC pin, plus the IOAPIC routing
1571 	 * entry is masked/unmasked at a high rate as well then sooner or
1572 	 * later IOAPIC line gets 'stuck', no more interrupts are received
1573 	 * from the device. If focus CPU is disabled then the hang goes
1574 	 * away, oh well :-(
1575 	 *
1576 	 * [ This bug can be reproduced easily with a level-triggered
1577 	 *   PCI Ne2000 networking cards and PII/PIII processors, dual
1578 	 *   BX chipset. ]
1579 	 */
1580 	/*
1581 	 * Actually disabling the focus CPU check just makes the hang less
1582 	 * frequent as it makes the interrupt distribution model be more
1583 	 * like LRU than MRU (the short-term load is more even across CPUs).
1584 	 */
1585 
1586 	/*
1587 	 * - enable focus processor (bit==0)
1588 	 * - 64bit mode always use processor focus
1589 	 *   so no need to set it
1590 	 */
1591 	value &= ~APIC_SPIV_FOCUS_DISABLED;
1592 #endif
1593 
1594 	/*
1595 	 * Set spurious IRQ vector
1596 	 */
1597 	value |= SPURIOUS_APIC_VECTOR;
1598 	apic_write(APIC_SPIV, value);
1599 
1600 	perf_events_lapic_init();
1601 
1602 	/*
1603 	 * Set up LVT0, LVT1:
1604 	 *
1605 	 * set up through-local-APIC on the boot CPU's LINT0. This is not
1606 	 * strictly necessary in pure symmetric-IO mode, but sometimes
1607 	 * we delegate interrupts to the 8259A.
1608 	 */
1609 	/*
1610 	 * TODO: set up through-local-APIC from through-I/O-APIC? --macro
1611 	 */
1612 	value = apic_read(APIC_LVT0) & APIC_LVT_MASKED;
1613 	if (!cpu && (pic_mode || !value || ioapic_is_disabled)) {
1614 		value = APIC_DM_EXTINT;
1615 		apic_pr_verbose("Enabled ExtINT on CPU#%d\n", cpu);
1616 	} else {
1617 		value = APIC_DM_EXTINT | APIC_LVT_MASKED;
1618 		apic_pr_verbose("Masked ExtINT on CPU#%d\n", cpu);
1619 	}
1620 	apic_write(APIC_LVT0, value);
1621 
1622 	/*
1623 	 * Only the BSP sees the LINT1 NMI signal by default. This can be
1624 	 * modified by apic_extnmi= boot option.
1625 	 */
1626 	if ((!cpu && apic_extnmi != APIC_EXTNMI_NONE) ||
1627 	    apic_extnmi == APIC_EXTNMI_ALL)
1628 		value = APIC_DM_NMI;
1629 	else
1630 		value = APIC_DM_NMI | APIC_LVT_MASKED;
1631 
1632 	/* Is 82489DX ? */
1633 	if (!lapic_is_integrated())
1634 		value |= APIC_LVT_LEVEL_TRIGGER;
1635 	apic_write(APIC_LVT1, value);
1636 
1637 #ifdef CONFIG_X86_MCE_INTEL
1638 	/* Recheck CMCI information after local APIC is up on CPU #0 */
1639 	if (!cpu)
1640 		cmci_recheck();
1641 #endif
1642 }
1643 
1644 static void end_local_APIC_setup(void)
1645 {
1646 	lapic_setup_esr();
1647 
1648 #ifdef CONFIG_X86_32
1649 	{
1650 		unsigned int value;
1651 		/* Disable the local apic timer */
1652 		value = apic_read(APIC_LVTT);
1653 		value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
1654 		apic_write(APIC_LVTT, value);
1655 	}
1656 #endif
1657 
1658 	apic_pm_activate();
1659 }
1660 
1661 /*
1662  * APIC setup function for application processors. Called from smpboot.c
1663  */
1664 void apic_ap_setup(void)
1665 {
1666 	setup_local_APIC();
1667 	end_local_APIC_setup();
1668 }
1669 
1670 static __init void apic_read_boot_cpu_id(bool x2apic)
1671 {
1672 	/*
1673 	 * This can be invoked from check_x2apic() before the APIC has been
1674 	 * selected. But that code knows for sure that the BIOS enabled
1675 	 * X2APIC.
1676 	 */
1677 	if (x2apic) {
1678 		boot_cpu_physical_apicid = native_apic_msr_read(APIC_ID);
1679 		boot_cpu_apic_version = GET_APIC_VERSION(native_apic_msr_read(APIC_LVR));
1680 	} else {
1681 		boot_cpu_physical_apicid = read_apic_id();
1682 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
1683 	}
1684 	topology_register_boot_apic(boot_cpu_physical_apicid);
1685 }
1686 
1687 #ifdef CONFIG_X86_X2APIC
1688 int x2apic_mode;
1689 EXPORT_SYMBOL_GPL(x2apic_mode);
1690 
1691 enum {
1692 	X2APIC_OFF,
1693 	X2APIC_DISABLED,
1694 	/* All states below here have X2APIC enabled */
1695 	X2APIC_ON,
1696 	X2APIC_ON_LOCKED
1697 };
1698 static int x2apic_state;
1699 
1700 static bool x2apic_hw_locked(void)
1701 {
1702 	u64 x86_arch_cap_msr;
1703 	u64 msr;
1704 
1705 	x86_arch_cap_msr = x86_read_arch_cap_msr();
1706 	if (x86_arch_cap_msr & ARCH_CAP_XAPIC_DISABLE) {
1707 		rdmsrq(MSR_IA32_XAPIC_DISABLE_STATUS, msr);
1708 		return (msr & LEGACY_XAPIC_DISABLED);
1709 	}
1710 	return false;
1711 }
1712 
1713 static void __x2apic_disable(void)
1714 {
1715 	u64 msr;
1716 
1717 	if (!boot_cpu_has(X86_FEATURE_APIC))
1718 		return;
1719 
1720 	rdmsrq(MSR_IA32_APICBASE, msr);
1721 	if (!(msr & X2APIC_ENABLE))
1722 		return;
1723 	/* Disable xapic and x2apic first and then reenable xapic mode */
1724 	wrmsrq(MSR_IA32_APICBASE, msr & ~(X2APIC_ENABLE | XAPIC_ENABLE));
1725 	wrmsrq(MSR_IA32_APICBASE, msr & ~X2APIC_ENABLE);
1726 	printk_once(KERN_INFO "x2apic disabled\n");
1727 }
1728 
1729 static void __x2apic_enable(void)
1730 {
1731 	u64 msr;
1732 
1733 	rdmsrq(MSR_IA32_APICBASE, msr);
1734 	if (msr & X2APIC_ENABLE)
1735 		return;
1736 	wrmsrq(MSR_IA32_APICBASE, msr | X2APIC_ENABLE);
1737 	printk_once(KERN_INFO "x2apic enabled\n");
1738 }
1739 
1740 static int __init setup_nox2apic(char *str)
1741 {
1742 	if (x2apic_enabled()) {
1743 		u32 apicid = native_apic_msr_read(APIC_ID);
1744 
1745 		if (apicid >= 255) {
1746 			pr_warn("Apicid: %08x, cannot enforce nox2apic\n",
1747 				apicid);
1748 			return 0;
1749 		}
1750 		if (x2apic_hw_locked()) {
1751 			pr_warn("APIC locked in x2apic mode, can't disable\n");
1752 			return 0;
1753 		}
1754 		pr_warn("x2apic already enabled.\n");
1755 		__x2apic_disable();
1756 	}
1757 	setup_clear_cpu_cap(X86_FEATURE_X2APIC);
1758 	x2apic_state = X2APIC_DISABLED;
1759 	x2apic_mode = 0;
1760 	return 0;
1761 }
1762 early_param("nox2apic", setup_nox2apic);
1763 
1764 /* Called from cpu_init() to enable x2apic on (secondary) cpus */
1765 void x2apic_setup(void)
1766 {
1767 	/*
1768 	 * Try to make the AP's APIC state match that of the BSP,  but if the
1769 	 * BSP is unlocked and the AP is locked then there is a state mismatch.
1770 	 * Warn about the mismatch in case a GP fault occurs due to a locked AP
1771 	 * trying to be turned off.
1772 	 */
1773 	if (x2apic_state != X2APIC_ON_LOCKED && x2apic_hw_locked())
1774 		pr_warn("x2apic lock mismatch between BSP and AP.\n");
1775 	/*
1776 	 * If x2apic is not in ON or LOCKED state, disable it if already enabled
1777 	 * from BIOS.
1778 	 */
1779 	if (x2apic_state < X2APIC_ON) {
1780 		__x2apic_disable();
1781 		return;
1782 	}
1783 	__x2apic_enable();
1784 }
1785 
1786 static __init void apic_set_fixmap(bool read_apic);
1787 
1788 static __init void x2apic_disable(void)
1789 {
1790 	u32 x2apic_id;
1791 
1792 	if (x2apic_state < X2APIC_ON)
1793 		return;
1794 
1795 	x2apic_id = read_apic_id();
1796 	if (x2apic_id >= 255)
1797 		panic("Cannot disable x2apic, id: %08x\n", x2apic_id);
1798 
1799 	if (x2apic_hw_locked()) {
1800 		pr_warn("Cannot disable locked x2apic, id: %08x\n", x2apic_id);
1801 		return;
1802 	}
1803 
1804 	__x2apic_disable();
1805 
1806 	x2apic_mode = 0;
1807 	x2apic_state = X2APIC_DISABLED;
1808 
1809 	/*
1810 	 * Don't reread the APIC ID as it was already done from
1811 	 * check_x2apic() and the APIC driver still is a x2APIC variant,
1812 	 * which fails to do the read after x2APIC was disabled.
1813 	 */
1814 	apic_set_fixmap(false);
1815 }
1816 
1817 static __init void x2apic_enable(void)
1818 {
1819 	if (x2apic_state != X2APIC_OFF)
1820 		return;
1821 
1822 	x2apic_mode = 1;
1823 	x2apic_state = X2APIC_ON;
1824 	__x2apic_enable();
1825 }
1826 
1827 static __init void try_to_enable_x2apic(int remap_mode)
1828 {
1829 	if (x2apic_state == X2APIC_DISABLED)
1830 		return;
1831 
1832 	if (remap_mode != IRQ_REMAP_X2APIC_MODE) {
1833 		u32 apic_limit = 255;
1834 
1835 		/*
1836 		 * Using X2APIC without IR is not architecturally supported
1837 		 * on bare metal but may be supported in guests.
1838 		 */
1839 		if (!x86_init.hyper.x2apic_available()) {
1840 			pr_info("x2apic: IRQ remapping doesn't support X2APIC mode\n");
1841 			x2apic_disable();
1842 			return;
1843 		}
1844 
1845 		/*
1846 		 * If the hypervisor supports extended destination ID in
1847 		 * MSI, that increases the maximum APIC ID that can be
1848 		 * used for non-remapped IRQ domains.
1849 		 */
1850 		if (x86_init.hyper.msi_ext_dest_id()) {
1851 			virt_ext_dest_id = 1;
1852 			apic_limit = 32767;
1853 		}
1854 
1855 		/*
1856 		 * Without IR, all CPUs can be addressed by IOAPIC/MSI only
1857 		 * in physical mode, and CPUs with an APIC ID that cannot
1858 		 * be addressed must not be brought online.
1859 		 */
1860 		x2apic_set_max_apicid(apic_limit);
1861 		x2apic_phys = 1;
1862 	}
1863 	x2apic_enable();
1864 }
1865 
1866 void __init check_x2apic(void)
1867 {
1868 	if (x2apic_enabled()) {
1869 		pr_info("x2apic: enabled by BIOS, switching to x2apic ops\n");
1870 		x2apic_mode = 1;
1871 		if (x2apic_hw_locked())
1872 			x2apic_state = X2APIC_ON_LOCKED;
1873 		else
1874 			x2apic_state = X2APIC_ON;
1875 		apic_read_boot_cpu_id(true);
1876 	} else if (!boot_cpu_has(X86_FEATURE_X2APIC)) {
1877 		x2apic_state = X2APIC_DISABLED;
1878 	}
1879 }
1880 #else /* CONFIG_X86_X2APIC */
1881 void __init check_x2apic(void)
1882 {
1883 	if (!apic_is_x2apic_enabled())
1884 		return;
1885 	/*
1886 	 * Checkme: Can we simply turn off x2APIC here instead of disabling the APIC?
1887 	 */
1888 	pr_err("Kernel does not support x2APIC, please recompile with CONFIG_X86_X2APIC.\n");
1889 	pr_err("Disabling APIC, expect reduced performance and functionality.\n");
1890 
1891 	apic_is_disabled = true;
1892 	setup_clear_cpu_cap(X86_FEATURE_APIC);
1893 }
1894 
1895 static inline void try_to_enable_x2apic(int remap_mode) { }
1896 static inline void __x2apic_enable(void) { }
1897 #endif /* !CONFIG_X86_X2APIC */
1898 
1899 void __init enable_IR_x2apic(void)
1900 {
1901 	unsigned long flags;
1902 	int ret, ir_stat;
1903 
1904 	if (ioapic_is_disabled) {
1905 		pr_info("Not enabling interrupt remapping due to skipped IO-APIC setup\n");
1906 		return;
1907 	}
1908 
1909 	ir_stat = irq_remapping_prepare();
1910 	if (ir_stat < 0 && !x2apic_supported())
1911 		return;
1912 
1913 	ret = save_ioapic_entries();
1914 	if (ret) {
1915 		pr_info("Saving IO-APIC state failed: %d\n", ret);
1916 		return;
1917 	}
1918 
1919 	local_irq_save(flags);
1920 	legacy_pic->mask_all();
1921 	mask_ioapic_entries();
1922 
1923 	/* If irq_remapping_prepare() succeeded, try to enable it */
1924 	if (ir_stat >= 0)
1925 		ir_stat = irq_remapping_enable();
1926 	/* ir_stat contains the remap mode or an error code */
1927 	try_to_enable_x2apic(ir_stat);
1928 
1929 	if (ir_stat < 0)
1930 		restore_ioapic_entries();
1931 	legacy_pic->restore_mask();
1932 	local_irq_restore(flags);
1933 }
1934 
1935 #ifdef CONFIG_X86_64
1936 /*
1937  * Detect and enable local APICs on non-SMP boards.
1938  * Original code written by Keir Fraser.
1939  * On AMD64 we trust the BIOS - if it says no APIC it is likely
1940  * not correctly set up (usually the APIC timer won't work etc.)
1941  */
1942 static bool __init detect_init_APIC(void)
1943 {
1944 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1945 		pr_info("No local APIC present\n");
1946 		return false;
1947 	}
1948 
1949 	register_lapic_address(APIC_DEFAULT_PHYS_BASE);
1950 	return true;
1951 }
1952 #else
1953 
1954 static bool __init apic_verify(unsigned long addr)
1955 {
1956 	u32 features, h, l;
1957 
1958 	/*
1959 	 * The APIC feature bit should now be enabled
1960 	 * in `cpuid'
1961 	 */
1962 	features = cpuid_edx(1);
1963 	if (!(features & (1 << X86_FEATURE_APIC))) {
1964 		pr_warn("Could not enable APIC!\n");
1965 		return false;
1966 	}
1967 	set_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
1968 
1969 	/* The BIOS may have set up the APIC at some other address */
1970 	if (boot_cpu_data.x86 >= 6) {
1971 		rdmsr(MSR_IA32_APICBASE, l, h);
1972 		if (l & MSR_IA32_APICBASE_ENABLE)
1973 			addr = l & MSR_IA32_APICBASE_BASE;
1974 	}
1975 
1976 	register_lapic_address(addr);
1977 	pr_info("Found and enabled local APIC!\n");
1978 	return true;
1979 }
1980 
1981 bool __init apic_force_enable(unsigned long addr)
1982 {
1983 	u32 h, l;
1984 
1985 	if (apic_is_disabled)
1986 		return false;
1987 
1988 	/*
1989 	 * Some BIOSes disable the local APIC in the APIC_BASE
1990 	 * MSR. This can only be done in software for Intel P6 or later
1991 	 * and AMD K7 (Model > 1) or later.
1992 	 */
1993 	if (boot_cpu_data.x86 >= 6) {
1994 		rdmsr(MSR_IA32_APICBASE, l, h);
1995 		if (!(l & MSR_IA32_APICBASE_ENABLE)) {
1996 			pr_info("Local APIC disabled by BIOS -- reenabling.\n");
1997 			l &= ~MSR_IA32_APICBASE_BASE;
1998 			l |= MSR_IA32_APICBASE_ENABLE | addr;
1999 			wrmsr(MSR_IA32_APICBASE, l, h);
2000 			enabled_via_apicbase = 1;
2001 		}
2002 	}
2003 	return apic_verify(addr);
2004 }
2005 
2006 /*
2007  * Detect and initialize APIC
2008  */
2009 static bool __init detect_init_APIC(void)
2010 {
2011 	/* Disabled by kernel option? */
2012 	if (apic_is_disabled)
2013 		return false;
2014 
2015 	switch (boot_cpu_data.x86_vendor) {
2016 	case X86_VENDOR_AMD:
2017 		if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model > 1) ||
2018 		    (boot_cpu_data.x86 >= 15))
2019 			break;
2020 		goto no_apic;
2021 	case X86_VENDOR_HYGON:
2022 		break;
2023 	case X86_VENDOR_INTEL:
2024 		if ((boot_cpu_data.x86 == 5 && boot_cpu_has(X86_FEATURE_APIC)) ||
2025 		    boot_cpu_data.x86_vfm >= INTEL_PENTIUM_PRO)
2026 			break;
2027 		goto no_apic;
2028 	default:
2029 		goto no_apic;
2030 	}
2031 
2032 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
2033 		/*
2034 		 * Over-ride BIOS and try to enable the local APIC only if
2035 		 * "lapic" specified.
2036 		 */
2037 		if (!force_enable_local_apic) {
2038 			pr_info("Local APIC disabled by BIOS -- "
2039 				"you can enable it with \"lapic\"\n");
2040 			return false;
2041 		}
2042 		if (!apic_force_enable(APIC_DEFAULT_PHYS_BASE))
2043 			return false;
2044 	} else {
2045 		if (!apic_verify(APIC_DEFAULT_PHYS_BASE))
2046 			return false;
2047 	}
2048 
2049 	apic_pm_activate();
2050 
2051 	return true;
2052 
2053 no_apic:
2054 	pr_info("No local APIC present or hardware disabled\n");
2055 	return false;
2056 }
2057 #endif
2058 
2059 /**
2060  * init_apic_mappings - initialize APIC mappings
2061  */
2062 void __init init_apic_mappings(void)
2063 {
2064 	if (apic_validate_deadline_timer())
2065 		pr_info("TSC deadline timer available\n");
2066 
2067 	if (x2apic_mode)
2068 		return;
2069 
2070 	if (!smp_found_config) {
2071 		if (!detect_init_APIC()) {
2072 			pr_info("APIC: disable apic facility\n");
2073 			apic_disable();
2074 		}
2075 	}
2076 }
2077 
2078 static __init void apic_set_fixmap(bool read_apic)
2079 {
2080 	set_fixmap_nocache(FIX_APIC_BASE, mp_lapic_addr);
2081 	apic_mmio_base = APIC_BASE;
2082 	apic_pr_verbose("Mapped APIC to %16lx (%16lx)\n", apic_mmio_base, mp_lapic_addr);
2083 	if (read_apic)
2084 		apic_read_boot_cpu_id(false);
2085 }
2086 
2087 void __init register_lapic_address(unsigned long address)
2088 {
2089 	/* This should only happen once */
2090 	WARN_ON_ONCE(mp_lapic_addr);
2091 	mp_lapic_addr = address;
2092 
2093 	if (!x2apic_mode)
2094 		apic_set_fixmap(true);
2095 }
2096 
2097 /*
2098  * Local APIC interrupts
2099  */
2100 
2101 /*
2102  * Common handling code for spurious_interrupt and spurious_vector entry
2103  * points below. No point in allowing the compiler to inline it twice.
2104  */
2105 static noinline void handle_spurious_interrupt(u8 vector)
2106 {
2107 	u32 v;
2108 
2109 	trace_spurious_apic_entry(vector);
2110 
2111 	inc_irq_stat(irq_spurious_count);
2112 
2113 	/*
2114 	 * If this is a spurious interrupt then do not acknowledge
2115 	 */
2116 	if (vector == SPURIOUS_APIC_VECTOR) {
2117 		/* See SDM vol 3 */
2118 		pr_info("Spurious APIC interrupt (vector 0xFF) on CPU#%d, should never happen.\n",
2119 			smp_processor_id());
2120 		goto out;
2121 	}
2122 
2123 	/*
2124 	 * If it is a vectored one, verify it's set in the ISR. If set,
2125 	 * acknowledge it.
2126 	 */
2127 	v = apic_read(APIC_ISR + ((vector & ~0x1f) >> 1));
2128 	if (v & (1 << (vector & 0x1f))) {
2129 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Acked\n",
2130 			vector, smp_processor_id());
2131 		apic_eoi();
2132 	} else {
2133 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Not pending!\n",
2134 			vector, smp_processor_id());
2135 	}
2136 out:
2137 	trace_spurious_apic_exit(vector);
2138 }
2139 
2140 /**
2141  * spurious_interrupt - Catch all for interrupts raised on unused vectors
2142  * @regs:	Pointer to pt_regs on stack
2143  * @vector:	The vector number
2144  *
2145  * This is invoked from ASM entry code to catch all interrupts which
2146  * trigger on an entry which is routed to the common_spurious idtentry
2147  * point.
2148  */
2149 DEFINE_IDTENTRY_IRQ(spurious_interrupt)
2150 {
2151 	handle_spurious_interrupt(vector);
2152 }
2153 
2154 DEFINE_IDTENTRY_SYSVEC(sysvec_spurious_apic_interrupt)
2155 {
2156 	handle_spurious_interrupt(SPURIOUS_APIC_VECTOR);
2157 }
2158 
2159 /*
2160  * This interrupt should never happen with our APIC/SMP architecture
2161  */
2162 DEFINE_IDTENTRY_SYSVEC(sysvec_error_interrupt)
2163 {
2164 	static const char * const error_interrupt_reason[] = {
2165 		"Send CS error",		/* APIC Error Bit 0 */
2166 		"Receive CS error",		/* APIC Error Bit 1 */
2167 		"Send accept error",		/* APIC Error Bit 2 */
2168 		"Receive accept error",		/* APIC Error Bit 3 */
2169 		"Redirectable IPI",		/* APIC Error Bit 4 */
2170 		"Send illegal vector",		/* APIC Error Bit 5 */
2171 		"Received illegal vector",	/* APIC Error Bit 6 */
2172 		"Illegal register address",	/* APIC Error Bit 7 */
2173 	};
2174 	u32 v, i = 0;
2175 
2176 	trace_error_apic_entry(ERROR_APIC_VECTOR);
2177 
2178 	/* First tickle the hardware, only then report what went on. -- REW */
2179 	if (lapic_get_maxlvt() > 3)	/* Due to the Pentium erratum 3AP. */
2180 		apic_write(APIC_ESR, 0);
2181 	v = apic_read(APIC_ESR);
2182 	apic_eoi();
2183 	atomic_inc(&irq_err_count);
2184 
2185 	apic_pr_debug("APIC error on CPU%d: %02x", smp_processor_id(), v);
2186 
2187 	v &= 0xff;
2188 	while (v) {
2189 		if (v & 0x1)
2190 			apic_pr_debug_cont(" : %s", error_interrupt_reason[i]);
2191 		i++;
2192 		v >>= 1;
2193 	}
2194 
2195 	apic_pr_debug_cont("\n");
2196 
2197 	trace_error_apic_exit(ERROR_APIC_VECTOR);
2198 }
2199 
2200 /**
2201  * connect_bsp_APIC - attach the APIC to the interrupt system
2202  */
2203 static void __init connect_bsp_APIC(void)
2204 {
2205 #ifdef CONFIG_X86_32
2206 	if (pic_mode) {
2207 		/*
2208 		 * Do not trust the local APIC being empty at bootup.
2209 		 */
2210 		clear_local_APIC();
2211 		/*
2212 		 * PIC mode, enable APIC mode in the IMCR, i.e.  connect BSP's
2213 		 * local APIC to INT and NMI lines.
2214 		 */
2215 		apic_pr_verbose("Leaving PIC mode, enabling APIC mode.\n");
2216 		imcr_pic_to_apic();
2217 	}
2218 #endif
2219 }
2220 
2221 /**
2222  * disconnect_bsp_APIC - detach the APIC from the interrupt system
2223  * @virt_wire_setup:	indicates, whether virtual wire mode is selected
2224  *
2225  * Virtual wire mode is necessary to deliver legacy interrupts even when the
2226  * APIC is disabled.
2227  */
2228 void disconnect_bsp_APIC(int virt_wire_setup)
2229 {
2230 	unsigned int value;
2231 
2232 #ifdef CONFIG_X86_32
2233 	if (pic_mode) {
2234 		/*
2235 		 * Put the board back into PIC mode (has an effect only on
2236 		 * certain older boards).  Note that APIC interrupts, including
2237 		 * IPIs, won't work beyond this point!  The only exception are
2238 		 * INIT IPIs.
2239 		 */
2240 		apic_pr_verbose("Disabling APIC mode, entering PIC mode.\n");
2241 		imcr_apic_to_pic();
2242 		return;
2243 	}
2244 #endif
2245 
2246 	/* Go back to Virtual Wire compatibility mode */
2247 
2248 	/* For the spurious interrupt use vector F, and enable it */
2249 	value = apic_read(APIC_SPIV);
2250 	value &= ~APIC_VECTOR_MASK;
2251 	value |= APIC_SPIV_APIC_ENABLED;
2252 	value |= 0xf;
2253 	apic_write(APIC_SPIV, value);
2254 
2255 	if (!virt_wire_setup) {
2256 		/*
2257 		 * For LVT0 make it edge triggered, active high,
2258 		 * external and enabled
2259 		 */
2260 		value = apic_read(APIC_LVT0);
2261 		value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2262 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2263 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2264 		value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2265 		value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_EXTINT);
2266 		apic_write(APIC_LVT0, value);
2267 	} else {
2268 		/* Disable LVT0 */
2269 		apic_write(APIC_LVT0, APIC_LVT_MASKED);
2270 	}
2271 
2272 	/*
2273 	 * For LVT1 make it edge triggered, active high,
2274 	 * nmi and enabled
2275 	 */
2276 	value = apic_read(APIC_LVT1);
2277 	value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2278 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2279 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2280 	value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2281 	value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_NMI);
2282 	apic_write(APIC_LVT1, value);
2283 }
2284 
2285 void __irq_msi_compose_msg(struct irq_cfg *cfg, struct msi_msg *msg,
2286 			   bool dmar)
2287 {
2288 	memset(msg, 0, sizeof(*msg));
2289 
2290 	msg->arch_addr_lo.base_address = X86_MSI_BASE_ADDRESS_LOW;
2291 	msg->arch_addr_lo.dest_mode_logical = apic->dest_mode_logical;
2292 	msg->arch_addr_lo.destid_0_7 = cfg->dest_apicid & 0xFF;
2293 
2294 	msg->arch_data.delivery_mode = APIC_DELIVERY_MODE_FIXED;
2295 	msg->arch_data.vector = cfg->vector;
2296 
2297 	msg->address_hi = X86_MSI_BASE_ADDRESS_HIGH;
2298 	/*
2299 	 * Only the IOMMU itself can use the trick of putting destination
2300 	 * APIC ID into the high bits of the address. Anything else would
2301 	 * just be writing to memory if it tried that, and needs IR to
2302 	 * address APICs which can't be addressed in the normal 32-bit
2303 	 * address range at 0xFFExxxxx. That is typically just 8 bits, but
2304 	 * some hypervisors allow the extended destination ID field in bits
2305 	 * 5-11 to be used, giving support for 15 bits of APIC IDs in total.
2306 	 */
2307 	if (dmar)
2308 		msg->arch_addr_hi.destid_8_31 = cfg->dest_apicid >> 8;
2309 	else if (virt_ext_dest_id && cfg->dest_apicid < 0x8000)
2310 		msg->arch_addr_lo.virt_destid_8_14 = cfg->dest_apicid >> 8;
2311 	else
2312 		WARN_ON_ONCE(cfg->dest_apicid > 0xFF);
2313 }
2314 
2315 u32 x86_msi_msg_get_destid(struct msi_msg *msg, bool extid)
2316 {
2317 	u32 dest = msg->arch_addr_lo.destid_0_7;
2318 
2319 	if (extid)
2320 		dest |= msg->arch_addr_hi.destid_8_31 << 8;
2321 	return dest;
2322 }
2323 EXPORT_SYMBOL_FOR_KVM(x86_msi_msg_get_destid);
2324 
2325 static void __init apic_bsp_up_setup(void)
2326 {
2327 	reset_phys_cpu_present_map(boot_cpu_physical_apicid);
2328 }
2329 
2330 /**
2331  * apic_bsp_setup - Setup function for local apic and io-apic
2332  * @upmode:		Force UP mode (for APIC_init_uniprocessor)
2333  */
2334 static void __init apic_bsp_setup(bool upmode)
2335 {
2336 	connect_bsp_APIC();
2337 	if (upmode)
2338 		apic_bsp_up_setup();
2339 	setup_local_APIC();
2340 
2341 	enable_IO_APIC();
2342 	end_local_APIC_setup();
2343 	irq_remap_enable_fault_handling();
2344 	setup_IO_APIC();
2345 	lapic_update_legacy_vectors();
2346 }
2347 
2348 #ifdef CONFIG_UP_LATE_INIT
2349 void __init up_late_init(void)
2350 {
2351 	if (apic_intr_mode == APIC_PIC)
2352 		return;
2353 
2354 	/* Setup local timer */
2355 	x86_init.timers.setup_percpu_clockev();
2356 }
2357 #endif
2358 
2359 /*
2360  * Power management
2361  */
2362 #ifdef CONFIG_PM
2363 
2364 static struct {
2365 	/*
2366 	 * 'active' is true if the local APIC was enabled by us and
2367 	 * not the BIOS; this signifies that we are also responsible
2368 	 * for disabling it before entering apm/acpi suspend
2369 	 */
2370 	int active;
2371 	/* r/w apic fields */
2372 	u32 apic_id;
2373 	unsigned int apic_taskpri;
2374 	unsigned int apic_ldr;
2375 	unsigned int apic_dfr;
2376 	unsigned int apic_spiv;
2377 	unsigned int apic_lvtt;
2378 	unsigned int apic_lvtpc;
2379 	unsigned int apic_lvt0;
2380 	unsigned int apic_lvt1;
2381 	unsigned int apic_lvterr;
2382 	unsigned int apic_tmict;
2383 	unsigned int apic_tdcr;
2384 	unsigned int apic_thmr;
2385 	unsigned int apic_cmci;
2386 } apic_pm_state;
2387 
2388 static int lapic_suspend(void)
2389 {
2390 	unsigned long flags;
2391 	int maxlvt;
2392 
2393 	if (!apic_pm_state.active)
2394 		return 0;
2395 
2396 	maxlvt = lapic_get_maxlvt();
2397 
2398 	apic_pm_state.apic_id = apic_read(APIC_ID);
2399 	apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
2400 	apic_pm_state.apic_ldr = apic_read(APIC_LDR);
2401 	apic_pm_state.apic_dfr = apic_read(APIC_DFR);
2402 	apic_pm_state.apic_spiv = apic_read(APIC_SPIV);
2403 	apic_pm_state.apic_lvtt = apic_read(APIC_LVTT);
2404 	if (maxlvt >= 4)
2405 		apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC);
2406 	apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0);
2407 	apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1);
2408 	apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR);
2409 	apic_pm_state.apic_tmict = apic_read(APIC_TMICT);
2410 	apic_pm_state.apic_tdcr = apic_read(APIC_TDCR);
2411 #ifdef CONFIG_X86_THERMAL_VECTOR
2412 	if (maxlvt >= 5)
2413 		apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
2414 #endif
2415 #ifdef CONFIG_X86_MCE_INTEL
2416 	if (maxlvt >= 6)
2417 		apic_pm_state.apic_cmci = apic_read(APIC_LVTCMCI);
2418 #endif
2419 
2420 	local_irq_save(flags);
2421 
2422 	/*
2423 	 * Mask IOAPIC before disabling the local APIC to prevent stale IRR
2424 	 * entries on some implementations.
2425 	 */
2426 	mask_ioapic_entries();
2427 
2428 	disable_local_APIC();
2429 
2430 	irq_remapping_disable();
2431 
2432 	local_irq_restore(flags);
2433 	return 0;
2434 }
2435 
2436 static void lapic_resume(void)
2437 {
2438 	unsigned int l, h;
2439 	unsigned long flags;
2440 	int maxlvt;
2441 
2442 	if (!apic_pm_state.active)
2443 		return;
2444 
2445 	local_irq_save(flags);
2446 
2447 	/*
2448 	 * IO-APIC and PIC have their own resume routines.
2449 	 * We just mask them here to make sure the interrupt
2450 	 * subsystem is completely quiet while we enable x2apic
2451 	 * and interrupt-remapping.
2452 	 */
2453 	mask_ioapic_entries();
2454 	legacy_pic->mask_all();
2455 
2456 	if (x2apic_mode) {
2457 		__x2apic_enable();
2458 	} else {
2459 		/*
2460 		 * Make sure the APICBASE points to the right address
2461 		 *
2462 		 * FIXME! This will be wrong if we ever support suspend on
2463 		 * SMP! We'll need to do this as part of the CPU restore!
2464 		 */
2465 		if (boot_cpu_data.x86 >= 6) {
2466 			rdmsr(MSR_IA32_APICBASE, l, h);
2467 			l &= ~MSR_IA32_APICBASE_BASE;
2468 			l |= MSR_IA32_APICBASE_ENABLE | mp_lapic_addr;
2469 			wrmsr(MSR_IA32_APICBASE, l, h);
2470 		}
2471 	}
2472 
2473 	maxlvt = lapic_get_maxlvt();
2474 	apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED);
2475 	apic_write(APIC_ID, apic_pm_state.apic_id);
2476 	apic_write(APIC_DFR, apic_pm_state.apic_dfr);
2477 	apic_write(APIC_LDR, apic_pm_state.apic_ldr);
2478 	apic_write(APIC_TASKPRI, apic_pm_state.apic_taskpri);
2479 	apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
2480 	apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
2481 	apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
2482 #ifdef CONFIG_X86_THERMAL_VECTOR
2483 	if (maxlvt >= 5)
2484 		apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
2485 #endif
2486 #ifdef CONFIG_X86_MCE_INTEL
2487 	if (maxlvt >= 6)
2488 		apic_write(APIC_LVTCMCI, apic_pm_state.apic_cmci);
2489 #endif
2490 	if (maxlvt >= 4)
2491 		apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
2492 	apic_write(APIC_LVTT, apic_pm_state.apic_lvtt);
2493 	apic_write(APIC_TDCR, apic_pm_state.apic_tdcr);
2494 	apic_write(APIC_TMICT, apic_pm_state.apic_tmict);
2495 	apic_write(APIC_ESR, 0);
2496 	apic_read(APIC_ESR);
2497 	apic_write(APIC_LVTERR, apic_pm_state.apic_lvterr);
2498 	apic_write(APIC_ESR, 0);
2499 	apic_read(APIC_ESR);
2500 
2501 	irq_remapping_reenable(x2apic_mode);
2502 
2503 	local_irq_restore(flags);
2504 }
2505 
2506 /*
2507  * This device has no shutdown method - fully functioning local APICs
2508  * are needed on every CPU up until machine_halt/restart/poweroff.
2509  */
2510 
2511 static struct syscore_ops lapic_syscore_ops = {
2512 	.resume		= lapic_resume,
2513 	.suspend	= lapic_suspend,
2514 };
2515 
2516 static void apic_pm_activate(void)
2517 {
2518 	apic_pm_state.active = 1;
2519 }
2520 
2521 static int __init init_lapic_sysfs(void)
2522 {
2523 	/* XXX: remove suspend/resume procs if !apic_pm_state.active? */
2524 	if (boot_cpu_has(X86_FEATURE_APIC))
2525 		register_syscore_ops(&lapic_syscore_ops);
2526 
2527 	return 0;
2528 }
2529 
2530 /* local apic needs to resume before other devices access its registers. */
2531 core_initcall(init_lapic_sysfs);
2532 
2533 #else	/* CONFIG_PM */
2534 
2535 static void apic_pm_activate(void) { }
2536 
2537 #endif	/* CONFIG_PM */
2538 
2539 #ifdef CONFIG_X86_64
2540 
2541 static int multi_checked;
2542 static int multi;
2543 
2544 static int set_multi(const struct dmi_system_id *d)
2545 {
2546 	if (multi)
2547 		return 0;
2548 	pr_info("APIC: %s detected, Multi Chassis\n", d->ident);
2549 	multi = 1;
2550 	return 0;
2551 }
2552 
2553 static const struct dmi_system_id multi_dmi_table[] = {
2554 	{
2555 		.callback = set_multi,
2556 		.ident = "IBM System Summit2",
2557 		.matches = {
2558 			DMI_MATCH(DMI_SYS_VENDOR, "IBM"),
2559 			DMI_MATCH(DMI_PRODUCT_NAME, "Summit2"),
2560 		},
2561 	},
2562 	{}
2563 };
2564 
2565 static void dmi_check_multi(void)
2566 {
2567 	if (multi_checked)
2568 		return;
2569 
2570 	dmi_check_system(multi_dmi_table);
2571 	multi_checked = 1;
2572 }
2573 
2574 /*
2575  * apic_is_clustered_box() -- Check if we can expect good TSC
2576  *
2577  * Thus far, the major user of this is IBM's Summit2 series:
2578  * Clustered boxes may have unsynced TSC problems if they are
2579  * multi-chassis.
2580  * Use DMI to check them
2581  */
2582 int apic_is_clustered_box(void)
2583 {
2584 	dmi_check_multi();
2585 	return multi;
2586 }
2587 #endif
2588 
2589 /*
2590  * APIC command line parameters
2591  */
2592 static int __init setup_nolapic(char *arg)
2593 {
2594 	apic_is_disabled = true;
2595 	setup_clear_cpu_cap(X86_FEATURE_APIC);
2596 	return 0;
2597 }
2598 early_param("nolapic", setup_nolapic);
2599 
2600 static int __init parse_lapic_timer_c2_ok(char *arg)
2601 {
2602 	local_apic_timer_c2_ok = 1;
2603 	return 0;
2604 }
2605 early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok);
2606 
2607 static int __init parse_disable_apic_timer(char *arg)
2608 {
2609 	disable_apic_timer = 1;
2610 	return 0;
2611 }
2612 early_param("noapictimer", parse_disable_apic_timer);
2613 
2614 static int __init parse_nolapic_timer(char *arg)
2615 {
2616 	disable_apic_timer = 1;
2617 	return 0;
2618 }
2619 early_param("nolapic_timer", parse_nolapic_timer);
2620 
2621 static int __init apic_set_verbosity(char *arg)
2622 {
2623 	if (!arg)  {
2624 		if (IS_ENABLED(CONFIG_X86_32))
2625 			return -EINVAL;
2626 
2627 		ioapic_is_disabled = false;
2628 		return 0;
2629 	}
2630 
2631 	if (strcmp("debug", arg) == 0)
2632 		apic_verbosity = APIC_DEBUG;
2633 	else if (strcmp("verbose", arg) == 0)
2634 		apic_verbosity = APIC_VERBOSE;
2635 #ifdef CONFIG_X86_64
2636 	else {
2637 		pr_warn("APIC Verbosity level %s not recognised"
2638 			" use apic=verbose or apic=debug\n", arg);
2639 		return -EINVAL;
2640 	}
2641 #endif
2642 
2643 	return 0;
2644 }
2645 early_param("apic", apic_set_verbosity);
2646 
2647 static int __init lapic_insert_resource(void)
2648 {
2649 	if (!apic_mmio_base)
2650 		return -1;
2651 
2652 	/* Put local APIC into the resource map. */
2653 	lapic_resource.start = apic_mmio_base;
2654 	lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
2655 	insert_resource(&iomem_resource, &lapic_resource);
2656 
2657 	return 0;
2658 }
2659 
2660 /*
2661  * need call insert after e820__reserve_resources()
2662  * that is using request_resource
2663  */
2664 late_initcall(lapic_insert_resource);
2665 
2666 static int __init apic_set_extnmi(char *arg)
2667 {
2668 	if (!arg)
2669 		return -EINVAL;
2670 
2671 	if (!strncmp("all", arg, 3))
2672 		apic_extnmi = APIC_EXTNMI_ALL;
2673 	else if (!strncmp("none", arg, 4))
2674 		apic_extnmi = APIC_EXTNMI_NONE;
2675 	else if (!strncmp("bsp", arg, 3))
2676 		apic_extnmi = APIC_EXTNMI_BSP;
2677 	else {
2678 		pr_warn("Unknown external NMI delivery mode `%s' ignored\n", arg);
2679 		return -EINVAL;
2680 	}
2681 
2682 	return 0;
2683 }
2684 early_param("apic_extnmi", apic_set_extnmi);
2685