xref: /openbmc/linux/arch/x86/kernel/apic/apic.c (revision 78bb17f7)
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/clockchips.h>
23 #include <linux/interrupt.h>
24 #include <linux/memblock.h>
25 #include <linux/ftrace.h>
26 #include <linux/ioport.h>
27 #include <linux/export.h>
28 #include <linux/syscore_ops.h>
29 #include <linux/delay.h>
30 #include <linux/timex.h>
31 #include <linux/i8253.h>
32 #include <linux/dmar.h>
33 #include <linux/init.h>
34 #include <linux/cpu.h>
35 #include <linux/dmi.h>
36 #include <linux/smp.h>
37 #include <linux/mm.h>
38 
39 #include <asm/trace/irq_vectors.h>
40 #include <asm/irq_remapping.h>
41 #include <asm/perf_event.h>
42 #include <asm/x86_init.h>
43 #include <asm/pgalloc.h>
44 #include <linux/atomic.h>
45 #include <asm/mpspec.h>
46 #include <asm/i8259.h>
47 #include <asm/proto.h>
48 #include <asm/traps.h>
49 #include <asm/apic.h>
50 #include <asm/io_apic.h>
51 #include <asm/desc.h>
52 #include <asm/hpet.h>
53 #include <asm/mtrr.h>
54 #include <asm/time.h>
55 #include <asm/smp.h>
56 #include <asm/mce.h>
57 #include <asm/tsc.h>
58 #include <asm/hypervisor.h>
59 #include <asm/cpu_device_id.h>
60 #include <asm/intel-family.h>
61 #include <asm/irq_regs.h>
62 
63 unsigned int num_processors;
64 
65 unsigned disabled_cpus;
66 
67 /* Processor that is doing the boot up */
68 unsigned int boot_cpu_physical_apicid __ro_after_init = -1U;
69 EXPORT_SYMBOL_GPL(boot_cpu_physical_apicid);
70 
71 u8 boot_cpu_apic_version __ro_after_init;
72 
73 /*
74  * The highest APIC ID seen during enumeration.
75  */
76 static unsigned int max_physical_apicid;
77 
78 /*
79  * Bitmask of physically existing CPUs:
80  */
81 physid_mask_t phys_cpu_present_map;
82 
83 /*
84  * Processor to be disabled specified by kernel parameter
85  * disable_cpu_apicid=<int>, mostly used for the kdump 2nd kernel to
86  * avoid undefined behaviour caused by sending INIT from AP to BSP.
87  */
88 static unsigned int disabled_cpu_apicid __ro_after_init = BAD_APICID;
89 
90 /*
91  * This variable controls which CPUs receive external NMIs.  By default,
92  * external NMIs are delivered only to the BSP.
93  */
94 static int apic_extnmi __ro_after_init = APIC_EXTNMI_BSP;
95 
96 /*
97  * Map cpu index to physical APIC ID
98  */
99 DEFINE_EARLY_PER_CPU_READ_MOSTLY(u16, x86_cpu_to_apicid, BAD_APICID);
100 DEFINE_EARLY_PER_CPU_READ_MOSTLY(u16, x86_bios_cpu_apicid, BAD_APICID);
101 DEFINE_EARLY_PER_CPU_READ_MOSTLY(u32, x86_cpu_to_acpiid, U32_MAX);
102 EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid);
103 EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
104 EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_acpiid);
105 
106 #ifdef CONFIG_X86_32
107 
108 /*
109  * On x86_32, the mapping between cpu and logical apicid may vary
110  * depending on apic in use.  The following early percpu variable is
111  * used for the mapping.  This is where the behaviors of x86_64 and 32
112  * actually diverge.  Let's keep it ugly for now.
113  */
114 DEFINE_EARLY_PER_CPU_READ_MOSTLY(int, x86_cpu_to_logical_apicid, BAD_APICID);
115 
116 /* Local APIC was disabled by the BIOS and enabled by the kernel */
117 static int enabled_via_apicbase __ro_after_init;
118 
119 /*
120  * Handle interrupt mode configuration register (IMCR).
121  * This register controls whether the interrupt signals
122  * that reach the BSP come from the master PIC or from the
123  * local APIC. Before entering Symmetric I/O Mode, either
124  * the BIOS or the operating system must switch out of
125  * PIC Mode by changing the IMCR.
126  */
127 static inline void imcr_pic_to_apic(void)
128 {
129 	/* select IMCR register */
130 	outb(0x70, 0x22);
131 	/* NMI and 8259 INTR go through APIC */
132 	outb(0x01, 0x23);
133 }
134 
135 static inline void imcr_apic_to_pic(void)
136 {
137 	/* select IMCR register */
138 	outb(0x70, 0x22);
139 	/* NMI and 8259 INTR go directly to BSP */
140 	outb(0x00, 0x23);
141 }
142 #endif
143 
144 /*
145  * Knob to control our willingness to enable the local APIC.
146  *
147  * +1=force-enable
148  */
149 static int force_enable_local_apic __initdata;
150 
151 /*
152  * APIC command line parameters
153  */
154 static int __init parse_lapic(char *arg)
155 {
156 	if (IS_ENABLED(CONFIG_X86_32) && !arg)
157 		force_enable_local_apic = 1;
158 	else if (arg && !strncmp(arg, "notscdeadline", 13))
159 		setup_clear_cpu_cap(X86_FEATURE_TSC_DEADLINE_TIMER);
160 	return 0;
161 }
162 early_param("lapic", parse_lapic);
163 
164 #ifdef CONFIG_X86_64
165 static int apic_calibrate_pmtmr __initdata;
166 static __init int setup_apicpmtimer(char *s)
167 {
168 	apic_calibrate_pmtmr = 1;
169 	notsc_setup(NULL);
170 	return 0;
171 }
172 __setup("apicpmtimer", setup_apicpmtimer);
173 #endif
174 
175 unsigned long mp_lapic_addr __ro_after_init;
176 int disable_apic __ro_after_init;
177 /* Disable local APIC timer from the kernel commandline or via dmi quirk */
178 static int disable_apic_timer __initdata;
179 /* Local APIC timer works in C2 */
180 int local_apic_timer_c2_ok __ro_after_init;
181 EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok);
182 
183 /*
184  * Debug level, exported for io_apic.c
185  */
186 int apic_verbosity __ro_after_init;
187 
188 int pic_mode __ro_after_init;
189 
190 /* Have we found an MP table */
191 int smp_found_config __ro_after_init;
192 
193 static struct resource lapic_resource = {
194 	.name = "Local APIC",
195 	.flags = IORESOURCE_MEM | IORESOURCE_BUSY,
196 };
197 
198 unsigned int lapic_timer_period = 0;
199 
200 static void apic_pm_activate(void);
201 
202 static unsigned long apic_phys __ro_after_init;
203 
204 /*
205  * Get the LAPIC version
206  */
207 static inline int lapic_get_version(void)
208 {
209 	return GET_APIC_VERSION(apic_read(APIC_LVR));
210 }
211 
212 /*
213  * Check, if the APIC is integrated or a separate chip
214  */
215 static inline int lapic_is_integrated(void)
216 {
217 	return APIC_INTEGRATED(lapic_get_version());
218 }
219 
220 /*
221  * Check, whether this is a modern or a first generation APIC
222  */
223 static int modern_apic(void)
224 {
225 	/* AMD systems use old APIC versions, so check the CPU */
226 	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
227 	    boot_cpu_data.x86 >= 0xf)
228 		return 1;
229 
230 	/* Hygon systems use modern APIC */
231 	if (boot_cpu_data.x86_vendor == X86_VENDOR_HYGON)
232 		return 1;
233 
234 	return lapic_get_version() >= 0x14;
235 }
236 
237 /*
238  * right after this call apic become NOOP driven
239  * so apic->write/read doesn't do anything
240  */
241 static void __init apic_disable(void)
242 {
243 	pr_info("APIC: switched to apic NOOP\n");
244 	apic = &apic_noop;
245 }
246 
247 void native_apic_wait_icr_idle(void)
248 {
249 	while (apic_read(APIC_ICR) & APIC_ICR_BUSY)
250 		cpu_relax();
251 }
252 
253 u32 native_safe_apic_wait_icr_idle(void)
254 {
255 	u32 send_status;
256 	int timeout;
257 
258 	timeout = 0;
259 	do {
260 		send_status = apic_read(APIC_ICR) & APIC_ICR_BUSY;
261 		if (!send_status)
262 			break;
263 		inc_irq_stat(icr_read_retry_count);
264 		udelay(100);
265 	} while (timeout++ < 1000);
266 
267 	return send_status;
268 }
269 
270 void native_apic_icr_write(u32 low, u32 id)
271 {
272 	unsigned long flags;
273 
274 	local_irq_save(flags);
275 	apic_write(APIC_ICR2, SET_APIC_DEST_FIELD(id));
276 	apic_write(APIC_ICR, low);
277 	local_irq_restore(flags);
278 }
279 
280 u64 native_apic_icr_read(void)
281 {
282 	u32 icr1, icr2;
283 
284 	icr2 = apic_read(APIC_ICR2);
285 	icr1 = apic_read(APIC_ICR);
286 
287 	return icr1 | ((u64)icr2 << 32);
288 }
289 
290 #ifdef CONFIG_X86_32
291 /**
292  * get_physical_broadcast - Get number of physical broadcast IDs
293  */
294 int get_physical_broadcast(void)
295 {
296 	return modern_apic() ? 0xff : 0xf;
297 }
298 #endif
299 
300 /**
301  * lapic_get_maxlvt - get the maximum number of local vector table entries
302  */
303 int lapic_get_maxlvt(void)
304 {
305 	/*
306 	 * - we always have APIC integrated on 64bit mode
307 	 * - 82489DXs do not report # of LVT entries
308 	 */
309 	return lapic_is_integrated() ? GET_APIC_MAXLVT(apic_read(APIC_LVR)) : 2;
310 }
311 
312 /*
313  * Local APIC timer
314  */
315 
316 /* Clock divisor */
317 #define APIC_DIVISOR 16
318 #define TSC_DIVISOR  8
319 
320 /*
321  * This function sets up the local APIC timer, with a timeout of
322  * 'clocks' APIC bus clock. During calibration we actually call
323  * this function twice on the boot CPU, once with a bogus timeout
324  * value, second time for real. The other (noncalibrating) CPUs
325  * call this function only once, with the real, calibrated value.
326  *
327  * We do reads before writes even if unnecessary, to get around the
328  * P5 APIC double write bug.
329  */
330 static void __setup_APIC_LVTT(unsigned int clocks, int oneshot, int irqen)
331 {
332 	unsigned int lvtt_value, tmp_value;
333 
334 	lvtt_value = LOCAL_TIMER_VECTOR;
335 	if (!oneshot)
336 		lvtt_value |= APIC_LVT_TIMER_PERIODIC;
337 	else if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
338 		lvtt_value |= APIC_LVT_TIMER_TSCDEADLINE;
339 
340 	if (!lapic_is_integrated())
341 		lvtt_value |= SET_APIC_TIMER_BASE(APIC_TIMER_BASE_DIV);
342 
343 	if (!irqen)
344 		lvtt_value |= APIC_LVT_MASKED;
345 
346 	apic_write(APIC_LVTT, lvtt_value);
347 
348 	if (lvtt_value & APIC_LVT_TIMER_TSCDEADLINE) {
349 		/*
350 		 * See Intel SDM: TSC-Deadline Mode chapter. In xAPIC mode,
351 		 * writing to the APIC LVTT and TSC_DEADLINE MSR isn't serialized.
352 		 * According to Intel, MFENCE can do the serialization here.
353 		 */
354 		asm volatile("mfence" : : : "memory");
355 		return;
356 	}
357 
358 	/*
359 	 * Divide PICLK by 16
360 	 */
361 	tmp_value = apic_read(APIC_TDCR);
362 	apic_write(APIC_TDCR,
363 		(tmp_value & ~(APIC_TDR_DIV_1 | APIC_TDR_DIV_TMBASE)) |
364 		APIC_TDR_DIV_16);
365 
366 	if (!oneshot)
367 		apic_write(APIC_TMICT, clocks / APIC_DIVISOR);
368 }
369 
370 /*
371  * Setup extended LVT, AMD specific
372  *
373  * Software should use the LVT offsets the BIOS provides.  The offsets
374  * are determined by the subsystems using it like those for MCE
375  * threshold or IBS.  On K8 only offset 0 (APIC500) and MCE interrupts
376  * are supported. Beginning with family 10h at least 4 offsets are
377  * available.
378  *
379  * Since the offsets must be consistent for all cores, we keep track
380  * of the LVT offsets in software and reserve the offset for the same
381  * vector also to be used on other cores. An offset is freed by
382  * setting the entry to APIC_EILVT_MASKED.
383  *
384  * If the BIOS is right, there should be no conflicts. Otherwise a
385  * "[Firmware Bug]: ..." error message is generated. However, if
386  * software does not properly determines the offsets, it is not
387  * necessarily a BIOS bug.
388  */
389 
390 static atomic_t eilvt_offsets[APIC_EILVT_NR_MAX];
391 
392 static inline int eilvt_entry_is_changeable(unsigned int old, unsigned int new)
393 {
394 	return (old & APIC_EILVT_MASKED)
395 		|| (new == APIC_EILVT_MASKED)
396 		|| ((new & ~APIC_EILVT_MASKED) == old);
397 }
398 
399 static unsigned int reserve_eilvt_offset(int offset, unsigned int new)
400 {
401 	unsigned int rsvd, vector;
402 
403 	if (offset >= APIC_EILVT_NR_MAX)
404 		return ~0;
405 
406 	rsvd = atomic_read(&eilvt_offsets[offset]);
407 	do {
408 		vector = rsvd & ~APIC_EILVT_MASKED;	/* 0: unassigned */
409 		if (vector && !eilvt_entry_is_changeable(vector, new))
410 			/* may not change if vectors are different */
411 			return rsvd;
412 		rsvd = atomic_cmpxchg(&eilvt_offsets[offset], rsvd, new);
413 	} while (rsvd != new);
414 
415 	rsvd &= ~APIC_EILVT_MASKED;
416 	if (rsvd && rsvd != vector)
417 		pr_info("LVT offset %d assigned for vector 0x%02x\n",
418 			offset, rsvd);
419 
420 	return new;
421 }
422 
423 /*
424  * If mask=1, the LVT entry does not generate interrupts while mask=0
425  * enables the vector. See also the BKDGs. Must be called with
426  * preemption disabled.
427  */
428 
429 int setup_APIC_eilvt(u8 offset, u8 vector, u8 msg_type, u8 mask)
430 {
431 	unsigned long reg = APIC_EILVTn(offset);
432 	unsigned int new, old, reserved;
433 
434 	new = (mask << 16) | (msg_type << 8) | vector;
435 	old = apic_read(reg);
436 	reserved = reserve_eilvt_offset(offset, new);
437 
438 	if (reserved != new) {
439 		pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
440 		       "vector 0x%x, but the register is already in use for "
441 		       "vector 0x%x on another cpu\n",
442 		       smp_processor_id(), reg, offset, new, reserved);
443 		return -EINVAL;
444 	}
445 
446 	if (!eilvt_entry_is_changeable(old, new)) {
447 		pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
448 		       "vector 0x%x, but the register is already in use for "
449 		       "vector 0x%x on this cpu\n",
450 		       smp_processor_id(), reg, offset, new, old);
451 		return -EBUSY;
452 	}
453 
454 	apic_write(reg, new);
455 
456 	return 0;
457 }
458 EXPORT_SYMBOL_GPL(setup_APIC_eilvt);
459 
460 /*
461  * Program the next event, relative to now
462  */
463 static int lapic_next_event(unsigned long delta,
464 			    struct clock_event_device *evt)
465 {
466 	apic_write(APIC_TMICT, delta);
467 	return 0;
468 }
469 
470 static int lapic_next_deadline(unsigned long delta,
471 			       struct clock_event_device *evt)
472 {
473 	u64 tsc;
474 
475 	tsc = rdtsc();
476 	wrmsrl(MSR_IA32_TSC_DEADLINE, tsc + (((u64) delta) * TSC_DIVISOR));
477 	return 0;
478 }
479 
480 static int lapic_timer_shutdown(struct clock_event_device *evt)
481 {
482 	unsigned int v;
483 
484 	/* Lapic used as dummy for broadcast ? */
485 	if (evt->features & CLOCK_EVT_FEAT_DUMMY)
486 		return 0;
487 
488 	v = apic_read(APIC_LVTT);
489 	v |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
490 	apic_write(APIC_LVTT, v);
491 	apic_write(APIC_TMICT, 0);
492 	return 0;
493 }
494 
495 static inline int
496 lapic_timer_set_periodic_oneshot(struct clock_event_device *evt, bool oneshot)
497 {
498 	/* Lapic used as dummy for broadcast ? */
499 	if (evt->features & CLOCK_EVT_FEAT_DUMMY)
500 		return 0;
501 
502 	__setup_APIC_LVTT(lapic_timer_period, oneshot, 1);
503 	return 0;
504 }
505 
506 static int lapic_timer_set_periodic(struct clock_event_device *evt)
507 {
508 	return lapic_timer_set_periodic_oneshot(evt, false);
509 }
510 
511 static int lapic_timer_set_oneshot(struct clock_event_device *evt)
512 {
513 	return lapic_timer_set_periodic_oneshot(evt, true);
514 }
515 
516 /*
517  * Local APIC timer broadcast function
518  */
519 static void lapic_timer_broadcast(const struct cpumask *mask)
520 {
521 #ifdef CONFIG_SMP
522 	apic->send_IPI_mask(mask, LOCAL_TIMER_VECTOR);
523 #endif
524 }
525 
526 
527 /*
528  * The local apic timer can be used for any function which is CPU local.
529  */
530 static struct clock_event_device lapic_clockevent = {
531 	.name				= "lapic",
532 	.features			= CLOCK_EVT_FEAT_PERIODIC |
533 					  CLOCK_EVT_FEAT_ONESHOT | CLOCK_EVT_FEAT_C3STOP
534 					  | CLOCK_EVT_FEAT_DUMMY,
535 	.shift				= 32,
536 	.set_state_shutdown		= lapic_timer_shutdown,
537 	.set_state_periodic		= lapic_timer_set_periodic,
538 	.set_state_oneshot		= lapic_timer_set_oneshot,
539 	.set_state_oneshot_stopped	= lapic_timer_shutdown,
540 	.set_next_event			= lapic_next_event,
541 	.broadcast			= lapic_timer_broadcast,
542 	.rating				= 100,
543 	.irq				= -1,
544 };
545 static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
546 
547 static __init u32 hsx_deadline_rev(void)
548 {
549 	switch (boot_cpu_data.x86_stepping) {
550 	case 0x02: return 0x3a; /* EP */
551 	case 0x04: return 0x0f; /* EX */
552 	}
553 
554 	return ~0U;
555 }
556 
557 static __init u32 bdx_deadline_rev(void)
558 {
559 	switch (boot_cpu_data.x86_stepping) {
560 	case 0x02: return 0x00000011;
561 	case 0x03: return 0x0700000e;
562 	case 0x04: return 0x0f00000c;
563 	case 0x05: return 0x0e000003;
564 	}
565 
566 	return ~0U;
567 }
568 
569 static __init u32 skx_deadline_rev(void)
570 {
571 	switch (boot_cpu_data.x86_stepping) {
572 	case 0x03: return 0x01000136;
573 	case 0x04: return 0x02000014;
574 	}
575 
576 	if (boot_cpu_data.x86_stepping > 4)
577 		return 0;
578 
579 	return ~0U;
580 }
581 
582 static const struct x86_cpu_id deadline_match[] __initconst = {
583 	X86_MATCH_INTEL_FAM6_MODEL( HASWELL_X,		&hsx_deadline_rev),
584 	X86_MATCH_INTEL_FAM6_MODEL( BROADWELL_X,	0x0b000020),
585 	X86_MATCH_INTEL_FAM6_MODEL( BROADWELL_D,	&bdx_deadline_rev),
586 	X86_MATCH_INTEL_FAM6_MODEL( SKYLAKE_X,		&skx_deadline_rev),
587 
588 	X86_MATCH_INTEL_FAM6_MODEL( HASWELL,		0x22),
589 	X86_MATCH_INTEL_FAM6_MODEL( HASWELL_L,		0x20),
590 	X86_MATCH_INTEL_FAM6_MODEL( HASWELL_G,		0x17),
591 
592 	X86_MATCH_INTEL_FAM6_MODEL( BROADWELL,		0x25),
593 	X86_MATCH_INTEL_FAM6_MODEL( BROADWELL_G,	0x17),
594 
595 	X86_MATCH_INTEL_FAM6_MODEL( SKYLAKE_L,		0xb2),
596 	X86_MATCH_INTEL_FAM6_MODEL( SKYLAKE,		0xb2),
597 
598 	X86_MATCH_INTEL_FAM6_MODEL( KABYLAKE_L,		0x52),
599 	X86_MATCH_INTEL_FAM6_MODEL( KABYLAKE,		0x52),
600 
601 	{},
602 };
603 
604 static __init bool apic_validate_deadline_timer(void)
605 {
606 	const struct x86_cpu_id *m;
607 	u32 rev;
608 
609 	if (!boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
610 		return false;
611 	if (boot_cpu_has(X86_FEATURE_HYPERVISOR))
612 		return true;
613 
614 	m = x86_match_cpu(deadline_match);
615 	if (!m)
616 		return true;
617 
618 	/*
619 	 * Function pointers will have the MSB set due to address layout,
620 	 * immediate revisions will not.
621 	 */
622 	if ((long)m->driver_data < 0)
623 		rev = ((u32 (*)(void))(m->driver_data))();
624 	else
625 		rev = (u32)m->driver_data;
626 
627 	if (boot_cpu_data.microcode >= rev)
628 		return true;
629 
630 	setup_clear_cpu_cap(X86_FEATURE_TSC_DEADLINE_TIMER);
631 	pr_err(FW_BUG "TSC_DEADLINE disabled due to Errata; "
632 	       "please update microcode to version: 0x%x (or later)\n", rev);
633 	return false;
634 }
635 
636 /*
637  * Setup the local APIC timer for this CPU. Copy the initialized values
638  * of the boot CPU and register the clock event in the framework.
639  */
640 static void setup_APIC_timer(void)
641 {
642 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
643 
644 	if (this_cpu_has(X86_FEATURE_ARAT)) {
645 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_C3STOP;
646 		/* Make LAPIC timer preferrable over percpu HPET */
647 		lapic_clockevent.rating = 150;
648 	}
649 
650 	memcpy(levt, &lapic_clockevent, sizeof(*levt));
651 	levt->cpumask = cpumask_of(smp_processor_id());
652 
653 	if (this_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER)) {
654 		levt->name = "lapic-deadline";
655 		levt->features &= ~(CLOCK_EVT_FEAT_PERIODIC |
656 				    CLOCK_EVT_FEAT_DUMMY);
657 		levt->set_next_event = lapic_next_deadline;
658 		clockevents_config_and_register(levt,
659 						tsc_khz * (1000 / TSC_DIVISOR),
660 						0xF, ~0UL);
661 	} else
662 		clockevents_register_device(levt);
663 }
664 
665 /*
666  * Install the updated TSC frequency from recalibration at the TSC
667  * deadline clockevent devices.
668  */
669 static void __lapic_update_tsc_freq(void *info)
670 {
671 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
672 
673 	if (!this_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
674 		return;
675 
676 	clockevents_update_freq(levt, tsc_khz * (1000 / TSC_DIVISOR));
677 }
678 
679 void lapic_update_tsc_freq(void)
680 {
681 	/*
682 	 * The clockevent device's ->mult and ->shift can both be
683 	 * changed. In order to avoid races, schedule the frequency
684 	 * update code on each CPU.
685 	 */
686 	on_each_cpu(__lapic_update_tsc_freq, NULL, 0);
687 }
688 
689 /*
690  * In this functions we calibrate APIC bus clocks to the external timer.
691  *
692  * We want to do the calibration only once since we want to have local timer
693  * irqs syncron. CPUs connected by the same APIC bus have the very same bus
694  * frequency.
695  *
696  * This was previously done by reading the PIT/HPET and waiting for a wrap
697  * around to find out, that a tick has elapsed. I have a box, where the PIT
698  * readout is broken, so it never gets out of the wait loop again. This was
699  * also reported by others.
700  *
701  * Monitoring the jiffies value is inaccurate and the clockevents
702  * infrastructure allows us to do a simple substitution of the interrupt
703  * handler.
704  *
705  * The calibration routine also uses the pm_timer when possible, as the PIT
706  * happens to run way too slow (factor 2.3 on my VAIO CoreDuo, which goes
707  * back to normal later in the boot process).
708  */
709 
710 #define LAPIC_CAL_LOOPS		(HZ/10)
711 
712 static __initdata int lapic_cal_loops = -1;
713 static __initdata long lapic_cal_t1, lapic_cal_t2;
714 static __initdata unsigned long long lapic_cal_tsc1, lapic_cal_tsc2;
715 static __initdata unsigned long lapic_cal_pm1, lapic_cal_pm2;
716 static __initdata unsigned long lapic_cal_j1, lapic_cal_j2;
717 
718 /*
719  * Temporary interrupt handler and polled calibration function.
720  */
721 static void __init lapic_cal_handler(struct clock_event_device *dev)
722 {
723 	unsigned long long tsc = 0;
724 	long tapic = apic_read(APIC_TMCCT);
725 	unsigned long pm = acpi_pm_read_early();
726 
727 	if (boot_cpu_has(X86_FEATURE_TSC))
728 		tsc = rdtsc();
729 
730 	switch (lapic_cal_loops++) {
731 	case 0:
732 		lapic_cal_t1 = tapic;
733 		lapic_cal_tsc1 = tsc;
734 		lapic_cal_pm1 = pm;
735 		lapic_cal_j1 = jiffies;
736 		break;
737 
738 	case LAPIC_CAL_LOOPS:
739 		lapic_cal_t2 = tapic;
740 		lapic_cal_tsc2 = tsc;
741 		if (pm < lapic_cal_pm1)
742 			pm += ACPI_PM_OVRRUN;
743 		lapic_cal_pm2 = pm;
744 		lapic_cal_j2 = jiffies;
745 		break;
746 	}
747 }
748 
749 static int __init
750 calibrate_by_pmtimer(long deltapm, long *delta, long *deltatsc)
751 {
752 	const long pm_100ms = PMTMR_TICKS_PER_SEC / 10;
753 	const long pm_thresh = pm_100ms / 100;
754 	unsigned long mult;
755 	u64 res;
756 
757 #ifndef CONFIG_X86_PM_TIMER
758 	return -1;
759 #endif
760 
761 	apic_printk(APIC_VERBOSE, "... PM-Timer delta = %ld\n", deltapm);
762 
763 	/* Check, if the PM timer is available */
764 	if (!deltapm)
765 		return -1;
766 
767 	mult = clocksource_hz2mult(PMTMR_TICKS_PER_SEC, 22);
768 
769 	if (deltapm > (pm_100ms - pm_thresh) &&
770 	    deltapm < (pm_100ms + pm_thresh)) {
771 		apic_printk(APIC_VERBOSE, "... PM-Timer result ok\n");
772 		return 0;
773 	}
774 
775 	res = (((u64)deltapm) *  mult) >> 22;
776 	do_div(res, 1000000);
777 	pr_warn("APIC calibration not consistent "
778 		"with PM-Timer: %ldms instead of 100ms\n", (long)res);
779 
780 	/* Correct the lapic counter value */
781 	res = (((u64)(*delta)) * pm_100ms);
782 	do_div(res, deltapm);
783 	pr_info("APIC delta adjusted to PM-Timer: "
784 		"%lu (%ld)\n", (unsigned long)res, *delta);
785 	*delta = (long)res;
786 
787 	/* Correct the tsc counter value */
788 	if (boot_cpu_has(X86_FEATURE_TSC)) {
789 		res = (((u64)(*deltatsc)) * pm_100ms);
790 		do_div(res, deltapm);
791 		apic_printk(APIC_VERBOSE, "TSC delta adjusted to "
792 					  "PM-Timer: %lu (%ld)\n",
793 					(unsigned long)res, *deltatsc);
794 		*deltatsc = (long)res;
795 	}
796 
797 	return 0;
798 }
799 
800 static int __init lapic_init_clockevent(void)
801 {
802 	if (!lapic_timer_period)
803 		return -1;
804 
805 	/* Calculate the scaled math multiplication factor */
806 	lapic_clockevent.mult = div_sc(lapic_timer_period/APIC_DIVISOR,
807 					TICK_NSEC, lapic_clockevent.shift);
808 	lapic_clockevent.max_delta_ns =
809 		clockevent_delta2ns(0x7FFFFFFF, &lapic_clockevent);
810 	lapic_clockevent.max_delta_ticks = 0x7FFFFFFF;
811 	lapic_clockevent.min_delta_ns =
812 		clockevent_delta2ns(0xF, &lapic_clockevent);
813 	lapic_clockevent.min_delta_ticks = 0xF;
814 
815 	return 0;
816 }
817 
818 bool __init apic_needs_pit(void)
819 {
820 	/*
821 	 * If the frequencies are not known, PIT is required for both TSC
822 	 * and apic timer calibration.
823 	 */
824 	if (!tsc_khz || !cpu_khz)
825 		return true;
826 
827 	/* Is there an APIC at all or is it disabled? */
828 	if (!boot_cpu_has(X86_FEATURE_APIC) || disable_apic)
829 		return true;
830 
831 	/*
832 	 * If interrupt delivery mode is legacy PIC or virtual wire without
833 	 * configuration, the local APIC timer wont be set up. Make sure
834 	 * that the PIT is initialized.
835 	 */
836 	if (apic_intr_mode == APIC_PIC ||
837 	    apic_intr_mode == APIC_VIRTUAL_WIRE_NO_CONFIG)
838 		return true;
839 
840 	/* Virt guests may lack ARAT, but still have DEADLINE */
841 	if (!boot_cpu_has(X86_FEATURE_ARAT))
842 		return true;
843 
844 	/* Deadline timer is based on TSC so no further PIT action required */
845 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
846 		return false;
847 
848 	/* APIC timer disabled? */
849 	if (disable_apic_timer)
850 		return true;
851 	/*
852 	 * The APIC timer frequency is known already, no PIT calibration
853 	 * required. If unknown, let the PIT be initialized.
854 	 */
855 	return lapic_timer_period == 0;
856 }
857 
858 static int __init calibrate_APIC_clock(void)
859 {
860 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
861 	u64 tsc_perj = 0, tsc_start = 0;
862 	unsigned long jif_start;
863 	unsigned long deltaj;
864 	long delta, deltatsc;
865 	int pm_referenced = 0;
866 
867 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
868 		return 0;
869 
870 	/*
871 	 * Check if lapic timer has already been calibrated by platform
872 	 * specific routine, such as tsc calibration code. If so just fill
873 	 * in the clockevent structure and return.
874 	 */
875 	if (!lapic_init_clockevent()) {
876 		apic_printk(APIC_VERBOSE, "lapic timer already calibrated %d\n",
877 			    lapic_timer_period);
878 		/*
879 		 * Direct calibration methods must have an always running
880 		 * local APIC timer, no need for broadcast timer.
881 		 */
882 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
883 		return 0;
884 	}
885 
886 	apic_printk(APIC_VERBOSE, "Using local APIC timer interrupts.\n"
887 		    "calibrating APIC timer ...\n");
888 
889 	/*
890 	 * There are platforms w/o global clockevent devices. Instead of
891 	 * making the calibration conditional on that, use a polling based
892 	 * approach everywhere.
893 	 */
894 	local_irq_disable();
895 
896 	/*
897 	 * Setup the APIC counter to maximum. There is no way the lapic
898 	 * can underflow in the 100ms detection time frame
899 	 */
900 	__setup_APIC_LVTT(0xffffffff, 0, 0);
901 
902 	/*
903 	 * Methods to terminate the calibration loop:
904 	 *  1) Global clockevent if available (jiffies)
905 	 *  2) TSC if available and frequency is known
906 	 */
907 	jif_start = READ_ONCE(jiffies);
908 
909 	if (tsc_khz) {
910 		tsc_start = rdtsc();
911 		tsc_perj = div_u64((u64)tsc_khz * 1000, HZ);
912 	}
913 
914 	/*
915 	 * Enable interrupts so the tick can fire, if a global
916 	 * clockevent device is available
917 	 */
918 	local_irq_enable();
919 
920 	while (lapic_cal_loops <= LAPIC_CAL_LOOPS) {
921 		/* Wait for a tick to elapse */
922 		while (1) {
923 			if (tsc_khz) {
924 				u64 tsc_now = rdtsc();
925 				if ((tsc_now - tsc_start) >= tsc_perj) {
926 					tsc_start += tsc_perj;
927 					break;
928 				}
929 			} else {
930 				unsigned long jif_now = READ_ONCE(jiffies);
931 
932 				if (time_after(jif_now, jif_start)) {
933 					jif_start = jif_now;
934 					break;
935 				}
936 			}
937 			cpu_relax();
938 		}
939 
940 		/* Invoke the calibration routine */
941 		local_irq_disable();
942 		lapic_cal_handler(NULL);
943 		local_irq_enable();
944 	}
945 
946 	local_irq_disable();
947 
948 	/* Build delta t1-t2 as apic timer counts down */
949 	delta = lapic_cal_t1 - lapic_cal_t2;
950 	apic_printk(APIC_VERBOSE, "... lapic delta = %ld\n", delta);
951 
952 	deltatsc = (long)(lapic_cal_tsc2 - lapic_cal_tsc1);
953 
954 	/* we trust the PM based calibration if possible */
955 	pm_referenced = !calibrate_by_pmtimer(lapic_cal_pm2 - lapic_cal_pm1,
956 					&delta, &deltatsc);
957 
958 	lapic_timer_period = (delta * APIC_DIVISOR) / LAPIC_CAL_LOOPS;
959 	lapic_init_clockevent();
960 
961 	apic_printk(APIC_VERBOSE, "..... delta %ld\n", delta);
962 	apic_printk(APIC_VERBOSE, "..... mult: %u\n", lapic_clockevent.mult);
963 	apic_printk(APIC_VERBOSE, "..... calibration result: %u\n",
964 		    lapic_timer_period);
965 
966 	if (boot_cpu_has(X86_FEATURE_TSC)) {
967 		apic_printk(APIC_VERBOSE, "..... CPU clock speed is "
968 			    "%ld.%04ld MHz.\n",
969 			    (deltatsc / LAPIC_CAL_LOOPS) / (1000000 / HZ),
970 			    (deltatsc / LAPIC_CAL_LOOPS) % (1000000 / HZ));
971 	}
972 
973 	apic_printk(APIC_VERBOSE, "..... host bus clock speed is "
974 		    "%u.%04u MHz.\n",
975 		    lapic_timer_period / (1000000 / HZ),
976 		    lapic_timer_period % (1000000 / HZ));
977 
978 	/*
979 	 * Do a sanity check on the APIC calibration result
980 	 */
981 	if (lapic_timer_period < (1000000 / HZ)) {
982 		local_irq_enable();
983 		pr_warn("APIC frequency too slow, disabling apic timer\n");
984 		return -1;
985 	}
986 
987 	levt->features &= ~CLOCK_EVT_FEAT_DUMMY;
988 
989 	/*
990 	 * PM timer calibration failed or not turned on so lets try APIC
991 	 * timer based calibration, if a global clockevent device is
992 	 * available.
993 	 */
994 	if (!pm_referenced && global_clock_event) {
995 		apic_printk(APIC_VERBOSE, "... verify APIC timer\n");
996 
997 		/*
998 		 * Setup the apic timer manually
999 		 */
1000 		levt->event_handler = lapic_cal_handler;
1001 		lapic_timer_set_periodic(levt);
1002 		lapic_cal_loops = -1;
1003 
1004 		/* Let the interrupts run */
1005 		local_irq_enable();
1006 
1007 		while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
1008 			cpu_relax();
1009 
1010 		/* Stop the lapic timer */
1011 		local_irq_disable();
1012 		lapic_timer_shutdown(levt);
1013 
1014 		/* Jiffies delta */
1015 		deltaj = lapic_cal_j2 - lapic_cal_j1;
1016 		apic_printk(APIC_VERBOSE, "... jiffies delta = %lu\n", deltaj);
1017 
1018 		/* Check, if the jiffies result is consistent */
1019 		if (deltaj >= LAPIC_CAL_LOOPS-2 && deltaj <= LAPIC_CAL_LOOPS+2)
1020 			apic_printk(APIC_VERBOSE, "... jiffies result ok\n");
1021 		else
1022 			levt->features |= CLOCK_EVT_FEAT_DUMMY;
1023 	}
1024 	local_irq_enable();
1025 
1026 	if (levt->features & CLOCK_EVT_FEAT_DUMMY) {
1027 		pr_warn("APIC timer disabled due to verification failure\n");
1028 		return -1;
1029 	}
1030 
1031 	return 0;
1032 }
1033 
1034 /*
1035  * Setup the boot APIC
1036  *
1037  * Calibrate and verify the result.
1038  */
1039 void __init setup_boot_APIC_clock(void)
1040 {
1041 	/*
1042 	 * The local apic timer can be disabled via the kernel
1043 	 * commandline or from the CPU detection code. Register the lapic
1044 	 * timer as a dummy clock event source on SMP systems, so the
1045 	 * broadcast mechanism is used. On UP systems simply ignore it.
1046 	 */
1047 	if (disable_apic_timer) {
1048 		pr_info("Disabling APIC timer\n");
1049 		/* No broadcast on UP ! */
1050 		if (num_possible_cpus() > 1) {
1051 			lapic_clockevent.mult = 1;
1052 			setup_APIC_timer();
1053 		}
1054 		return;
1055 	}
1056 
1057 	if (calibrate_APIC_clock()) {
1058 		/* No broadcast on UP ! */
1059 		if (num_possible_cpus() > 1)
1060 			setup_APIC_timer();
1061 		return;
1062 	}
1063 
1064 	/*
1065 	 * If nmi_watchdog is set to IO_APIC, we need the
1066 	 * PIT/HPET going.  Otherwise register lapic as a dummy
1067 	 * device.
1068 	 */
1069 	lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
1070 
1071 	/* Setup the lapic or request the broadcast */
1072 	setup_APIC_timer();
1073 	amd_e400_c1e_apic_setup();
1074 }
1075 
1076 void setup_secondary_APIC_clock(void)
1077 {
1078 	setup_APIC_timer();
1079 	amd_e400_c1e_apic_setup();
1080 }
1081 
1082 /*
1083  * The guts of the apic timer interrupt
1084  */
1085 static void local_apic_timer_interrupt(void)
1086 {
1087 	struct clock_event_device *evt = this_cpu_ptr(&lapic_events);
1088 
1089 	/*
1090 	 * Normally we should not be here till LAPIC has been initialized but
1091 	 * in some cases like kdump, its possible that there is a pending LAPIC
1092 	 * timer interrupt from previous kernel's context and is delivered in
1093 	 * new kernel the moment interrupts are enabled.
1094 	 *
1095 	 * Interrupts are enabled early and LAPIC is setup much later, hence
1096 	 * its possible that when we get here evt->event_handler is NULL.
1097 	 * Check for event_handler being NULL and discard the interrupt as
1098 	 * spurious.
1099 	 */
1100 	if (!evt->event_handler) {
1101 		pr_warn("Spurious LAPIC timer interrupt on cpu %d\n",
1102 			smp_processor_id());
1103 		/* Switch it off */
1104 		lapic_timer_shutdown(evt);
1105 		return;
1106 	}
1107 
1108 	/*
1109 	 * the NMI deadlock-detector uses this.
1110 	 */
1111 	inc_irq_stat(apic_timer_irqs);
1112 
1113 	evt->event_handler(evt);
1114 }
1115 
1116 /*
1117  * Local APIC timer interrupt. This is the most natural way for doing
1118  * local interrupts, but local timer interrupts can be emulated by
1119  * broadcast interrupts too. [in case the hw doesn't support APIC timers]
1120  *
1121  * [ if a single-CPU system runs an SMP kernel then we call the local
1122  *   interrupt as well. Thus we cannot inline the local irq ... ]
1123  */
1124 __visible void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
1125 {
1126 	struct pt_regs *old_regs = set_irq_regs(regs);
1127 
1128 	/*
1129 	 * NOTE! We'd better ACK the irq immediately,
1130 	 * because timer handling can be slow.
1131 	 *
1132 	 * update_process_times() expects us to have done irq_enter().
1133 	 * Besides, if we don't timer interrupts ignore the global
1134 	 * interrupt lock, which is the WrongThing (tm) to do.
1135 	 */
1136 	entering_ack_irq();
1137 	trace_local_timer_entry(LOCAL_TIMER_VECTOR);
1138 	local_apic_timer_interrupt();
1139 	trace_local_timer_exit(LOCAL_TIMER_VECTOR);
1140 	exiting_irq();
1141 
1142 	set_irq_regs(old_regs);
1143 }
1144 
1145 int setup_profiling_timer(unsigned int multiplier)
1146 {
1147 	return -EINVAL;
1148 }
1149 
1150 /*
1151  * Local APIC start and shutdown
1152  */
1153 
1154 /**
1155  * clear_local_APIC - shutdown the local APIC
1156  *
1157  * This is called, when a CPU is disabled and before rebooting, so the state of
1158  * the local APIC has no dangling leftovers. Also used to cleanout any BIOS
1159  * leftovers during boot.
1160  */
1161 void clear_local_APIC(void)
1162 {
1163 	int maxlvt;
1164 	u32 v;
1165 
1166 	/* APIC hasn't been mapped yet */
1167 	if (!x2apic_mode && !apic_phys)
1168 		return;
1169 
1170 	maxlvt = lapic_get_maxlvt();
1171 	/*
1172 	 * Masking an LVT entry can trigger a local APIC error
1173 	 * if the vector is zero. Mask LVTERR first to prevent this.
1174 	 */
1175 	if (maxlvt >= 3) {
1176 		v = ERROR_APIC_VECTOR; /* any non-zero vector will do */
1177 		apic_write(APIC_LVTERR, v | APIC_LVT_MASKED);
1178 	}
1179 	/*
1180 	 * Careful: we have to set masks only first to deassert
1181 	 * any level-triggered sources.
1182 	 */
1183 	v = apic_read(APIC_LVTT);
1184 	apic_write(APIC_LVTT, v | APIC_LVT_MASKED);
1185 	v = apic_read(APIC_LVT0);
1186 	apic_write(APIC_LVT0, v | APIC_LVT_MASKED);
1187 	v = apic_read(APIC_LVT1);
1188 	apic_write(APIC_LVT1, v | APIC_LVT_MASKED);
1189 	if (maxlvt >= 4) {
1190 		v = apic_read(APIC_LVTPC);
1191 		apic_write(APIC_LVTPC, v | APIC_LVT_MASKED);
1192 	}
1193 
1194 	/* lets not touch this if we didn't frob it */
1195 #ifdef CONFIG_X86_THERMAL_VECTOR
1196 	if (maxlvt >= 5) {
1197 		v = apic_read(APIC_LVTTHMR);
1198 		apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED);
1199 	}
1200 #endif
1201 #ifdef CONFIG_X86_MCE_INTEL
1202 	if (maxlvt >= 6) {
1203 		v = apic_read(APIC_LVTCMCI);
1204 		if (!(v & APIC_LVT_MASKED))
1205 			apic_write(APIC_LVTCMCI, v | APIC_LVT_MASKED);
1206 	}
1207 #endif
1208 
1209 	/*
1210 	 * Clean APIC state for other OSs:
1211 	 */
1212 	apic_write(APIC_LVTT, APIC_LVT_MASKED);
1213 	apic_write(APIC_LVT0, APIC_LVT_MASKED);
1214 	apic_write(APIC_LVT1, APIC_LVT_MASKED);
1215 	if (maxlvt >= 3)
1216 		apic_write(APIC_LVTERR, APIC_LVT_MASKED);
1217 	if (maxlvt >= 4)
1218 		apic_write(APIC_LVTPC, APIC_LVT_MASKED);
1219 
1220 	/* Integrated APIC (!82489DX) ? */
1221 	if (lapic_is_integrated()) {
1222 		if (maxlvt > 3)
1223 			/* Clear ESR due to Pentium errata 3AP and 11AP */
1224 			apic_write(APIC_ESR, 0);
1225 		apic_read(APIC_ESR);
1226 	}
1227 }
1228 
1229 /**
1230  * apic_soft_disable - Clears and software disables the local APIC on hotplug
1231  *
1232  * Contrary to disable_local_APIC() this does not touch the enable bit in
1233  * MSR_IA32_APICBASE. Clearing that bit on systems based on the 3 wire APIC
1234  * bus would require a hardware reset as the APIC would lose track of bus
1235  * arbitration. On systems with FSB delivery APICBASE could be disabled,
1236  * but it has to be guaranteed that no interrupt is sent to the APIC while
1237  * in that state and it's not clear from the SDM whether it still responds
1238  * to INIT/SIPI messages. Stay on the safe side and use software disable.
1239  */
1240 void apic_soft_disable(void)
1241 {
1242 	u32 value;
1243 
1244 	clear_local_APIC();
1245 
1246 	/* Soft disable APIC (implies clearing of registers for 82489DX!). */
1247 	value = apic_read(APIC_SPIV);
1248 	value &= ~APIC_SPIV_APIC_ENABLED;
1249 	apic_write(APIC_SPIV, value);
1250 }
1251 
1252 /**
1253  * disable_local_APIC - clear and disable the local APIC
1254  */
1255 void disable_local_APIC(void)
1256 {
1257 	/* APIC hasn't been mapped yet */
1258 	if (!x2apic_mode && !apic_phys)
1259 		return;
1260 
1261 	apic_soft_disable();
1262 
1263 #ifdef CONFIG_X86_32
1264 	/*
1265 	 * When LAPIC was disabled by the BIOS and enabled by the kernel,
1266 	 * restore the disabled state.
1267 	 */
1268 	if (enabled_via_apicbase) {
1269 		unsigned int l, h;
1270 
1271 		rdmsr(MSR_IA32_APICBASE, l, h);
1272 		l &= ~MSR_IA32_APICBASE_ENABLE;
1273 		wrmsr(MSR_IA32_APICBASE, l, h);
1274 	}
1275 #endif
1276 }
1277 
1278 /*
1279  * If Linux enabled the LAPIC against the BIOS default disable it down before
1280  * re-entering the BIOS on shutdown.  Otherwise the BIOS may get confused and
1281  * not power-off.  Additionally clear all LVT entries before disable_local_APIC
1282  * for the case where Linux didn't enable the LAPIC.
1283  */
1284 void lapic_shutdown(void)
1285 {
1286 	unsigned long flags;
1287 
1288 	if (!boot_cpu_has(X86_FEATURE_APIC) && !apic_from_smp_config())
1289 		return;
1290 
1291 	local_irq_save(flags);
1292 
1293 #ifdef CONFIG_X86_32
1294 	if (!enabled_via_apicbase)
1295 		clear_local_APIC();
1296 	else
1297 #endif
1298 		disable_local_APIC();
1299 
1300 
1301 	local_irq_restore(flags);
1302 }
1303 
1304 /**
1305  * sync_Arb_IDs - synchronize APIC bus arbitration IDs
1306  */
1307 void __init sync_Arb_IDs(void)
1308 {
1309 	/*
1310 	 * Unsupported on P4 - see Intel Dev. Manual Vol. 3, Ch. 8.6.1 And not
1311 	 * needed on AMD.
1312 	 */
1313 	if (modern_apic() || boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
1314 		return;
1315 
1316 	/*
1317 	 * Wait for idle.
1318 	 */
1319 	apic_wait_icr_idle();
1320 
1321 	apic_printk(APIC_DEBUG, "Synchronizing Arb IDs.\n");
1322 	apic_write(APIC_ICR, APIC_DEST_ALLINC |
1323 			APIC_INT_LEVELTRIG | APIC_DM_INIT);
1324 }
1325 
1326 enum apic_intr_mode_id apic_intr_mode __ro_after_init;
1327 
1328 static int __init __apic_intr_mode_select(void)
1329 {
1330 	/* Check kernel option */
1331 	if (disable_apic) {
1332 		pr_info("APIC disabled via kernel command line\n");
1333 		return APIC_PIC;
1334 	}
1335 
1336 	/* Check BIOS */
1337 #ifdef CONFIG_X86_64
1338 	/* On 64-bit, the APIC must be integrated, Check local APIC only */
1339 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1340 		disable_apic = 1;
1341 		pr_info("APIC disabled by BIOS\n");
1342 		return APIC_PIC;
1343 	}
1344 #else
1345 	/* On 32-bit, the APIC may be integrated APIC or 82489DX */
1346 
1347 	/* Neither 82489DX nor integrated APIC ? */
1348 	if (!boot_cpu_has(X86_FEATURE_APIC) && !smp_found_config) {
1349 		disable_apic = 1;
1350 		return APIC_PIC;
1351 	}
1352 
1353 	/* If the BIOS pretends there is an integrated APIC ? */
1354 	if (!boot_cpu_has(X86_FEATURE_APIC) &&
1355 		APIC_INTEGRATED(boot_cpu_apic_version)) {
1356 		disable_apic = 1;
1357 		pr_err(FW_BUG "Local APIC %d not detected, force emulation\n",
1358 				       boot_cpu_physical_apicid);
1359 		return APIC_PIC;
1360 	}
1361 #endif
1362 
1363 	/* Check MP table or ACPI MADT configuration */
1364 	if (!smp_found_config) {
1365 		disable_ioapic_support();
1366 		if (!acpi_lapic) {
1367 			pr_info("APIC: ACPI MADT or MP tables are not detected\n");
1368 			return APIC_VIRTUAL_WIRE_NO_CONFIG;
1369 		}
1370 		return APIC_VIRTUAL_WIRE;
1371 	}
1372 
1373 #ifdef CONFIG_SMP
1374 	/* If SMP should be disabled, then really disable it! */
1375 	if (!setup_max_cpus) {
1376 		pr_info("APIC: SMP mode deactivated\n");
1377 		return APIC_SYMMETRIC_IO_NO_ROUTING;
1378 	}
1379 
1380 	if (read_apic_id() != boot_cpu_physical_apicid) {
1381 		panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
1382 		     read_apic_id(), boot_cpu_physical_apicid);
1383 		/* Or can we switch back to PIC here? */
1384 	}
1385 #endif
1386 
1387 	return APIC_SYMMETRIC_IO;
1388 }
1389 
1390 /* Select the interrupt delivery mode for the BSP */
1391 void __init apic_intr_mode_select(void)
1392 {
1393 	apic_intr_mode = __apic_intr_mode_select();
1394 }
1395 
1396 /*
1397  * An initial setup of the virtual wire mode.
1398  */
1399 void __init init_bsp_APIC(void)
1400 {
1401 	unsigned int value;
1402 
1403 	/*
1404 	 * Don't do the setup now if we have a SMP BIOS as the
1405 	 * through-I/O-APIC virtual wire mode might be active.
1406 	 */
1407 	if (smp_found_config || !boot_cpu_has(X86_FEATURE_APIC))
1408 		return;
1409 
1410 	/*
1411 	 * Do not trust the local APIC being empty at bootup.
1412 	 */
1413 	clear_local_APIC();
1414 
1415 	/*
1416 	 * Enable APIC.
1417 	 */
1418 	value = apic_read(APIC_SPIV);
1419 	value &= ~APIC_VECTOR_MASK;
1420 	value |= APIC_SPIV_APIC_ENABLED;
1421 
1422 #ifdef CONFIG_X86_32
1423 	/* This bit is reserved on P4/Xeon and should be cleared */
1424 	if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
1425 	    (boot_cpu_data.x86 == 15))
1426 		value &= ~APIC_SPIV_FOCUS_DISABLED;
1427 	else
1428 #endif
1429 		value |= APIC_SPIV_FOCUS_DISABLED;
1430 	value |= SPURIOUS_APIC_VECTOR;
1431 	apic_write(APIC_SPIV, value);
1432 
1433 	/*
1434 	 * Set up the virtual wire mode.
1435 	 */
1436 	apic_write(APIC_LVT0, APIC_DM_EXTINT);
1437 	value = APIC_DM_NMI;
1438 	if (!lapic_is_integrated())		/* 82489DX */
1439 		value |= APIC_LVT_LEVEL_TRIGGER;
1440 	if (apic_extnmi == APIC_EXTNMI_NONE)
1441 		value |= APIC_LVT_MASKED;
1442 	apic_write(APIC_LVT1, value);
1443 }
1444 
1445 static void __init apic_bsp_setup(bool upmode);
1446 
1447 /* Init the interrupt delivery mode for the BSP */
1448 void __init apic_intr_mode_init(void)
1449 {
1450 	bool upmode = IS_ENABLED(CONFIG_UP_LATE_INIT);
1451 
1452 	switch (apic_intr_mode) {
1453 	case APIC_PIC:
1454 		pr_info("APIC: Keep in PIC mode(8259)\n");
1455 		return;
1456 	case APIC_VIRTUAL_WIRE:
1457 		pr_info("APIC: Switch to virtual wire mode setup\n");
1458 		default_setup_apic_routing();
1459 		break;
1460 	case APIC_VIRTUAL_WIRE_NO_CONFIG:
1461 		pr_info("APIC: Switch to virtual wire mode setup with no configuration\n");
1462 		upmode = true;
1463 		default_setup_apic_routing();
1464 		break;
1465 	case APIC_SYMMETRIC_IO:
1466 		pr_info("APIC: Switch to symmetric I/O mode setup\n");
1467 		default_setup_apic_routing();
1468 		break;
1469 	case APIC_SYMMETRIC_IO_NO_ROUTING:
1470 		pr_info("APIC: Switch to symmetric I/O mode setup in no SMP routine\n");
1471 		break;
1472 	}
1473 
1474 	apic_bsp_setup(upmode);
1475 }
1476 
1477 static void lapic_setup_esr(void)
1478 {
1479 	unsigned int oldvalue, value, maxlvt;
1480 
1481 	if (!lapic_is_integrated()) {
1482 		pr_info("No ESR for 82489DX.\n");
1483 		return;
1484 	}
1485 
1486 	if (apic->disable_esr) {
1487 		/*
1488 		 * Something untraceable is creating bad interrupts on
1489 		 * secondary quads ... for the moment, just leave the
1490 		 * ESR disabled - we can't do anything useful with the
1491 		 * errors anyway - mbligh
1492 		 */
1493 		pr_info("Leaving ESR disabled.\n");
1494 		return;
1495 	}
1496 
1497 	maxlvt = lapic_get_maxlvt();
1498 	if (maxlvt > 3)		/* Due to the Pentium erratum 3AP. */
1499 		apic_write(APIC_ESR, 0);
1500 	oldvalue = apic_read(APIC_ESR);
1501 
1502 	/* enables sending errors */
1503 	value = ERROR_APIC_VECTOR;
1504 	apic_write(APIC_LVTERR, value);
1505 
1506 	/*
1507 	 * spec says clear errors after enabling vector.
1508 	 */
1509 	if (maxlvt > 3)
1510 		apic_write(APIC_ESR, 0);
1511 	value = apic_read(APIC_ESR);
1512 	if (value != oldvalue)
1513 		apic_printk(APIC_VERBOSE, "ESR value before enabling "
1514 			"vector: 0x%08x  after: 0x%08x\n",
1515 			oldvalue, value);
1516 }
1517 
1518 #define APIC_IR_REGS		APIC_ISR_NR
1519 #define APIC_IR_BITS		(APIC_IR_REGS * 32)
1520 #define APIC_IR_MAPSIZE		(APIC_IR_BITS / BITS_PER_LONG)
1521 
1522 union apic_ir {
1523 	unsigned long	map[APIC_IR_MAPSIZE];
1524 	u32		regs[APIC_IR_REGS];
1525 };
1526 
1527 static bool apic_check_and_ack(union apic_ir *irr, union apic_ir *isr)
1528 {
1529 	int i, bit;
1530 
1531 	/* Read the IRRs */
1532 	for (i = 0; i < APIC_IR_REGS; i++)
1533 		irr->regs[i] = apic_read(APIC_IRR + i * 0x10);
1534 
1535 	/* Read the ISRs */
1536 	for (i = 0; i < APIC_IR_REGS; i++)
1537 		isr->regs[i] = apic_read(APIC_ISR + i * 0x10);
1538 
1539 	/*
1540 	 * If the ISR map is not empty. ACK the APIC and run another round
1541 	 * to verify whether a pending IRR has been unblocked and turned
1542 	 * into a ISR.
1543 	 */
1544 	if (!bitmap_empty(isr->map, APIC_IR_BITS)) {
1545 		/*
1546 		 * There can be multiple ISR bits set when a high priority
1547 		 * interrupt preempted a lower priority one. Issue an ACK
1548 		 * per set bit.
1549 		 */
1550 		for_each_set_bit(bit, isr->map, APIC_IR_BITS)
1551 			ack_APIC_irq();
1552 		return true;
1553 	}
1554 
1555 	return !bitmap_empty(irr->map, APIC_IR_BITS);
1556 }
1557 
1558 /*
1559  * After a crash, we no longer service the interrupts and a pending
1560  * interrupt from previous kernel might still have ISR bit set.
1561  *
1562  * Most probably by now the CPU has serviced that pending interrupt and it
1563  * might not have done the ack_APIC_irq() because it thought, interrupt
1564  * came from i8259 as ExtInt. LAPIC did not get EOI so it does not clear
1565  * the ISR bit and cpu thinks it has already serivced the interrupt. Hence
1566  * a vector might get locked. It was noticed for timer irq (vector
1567  * 0x31). Issue an extra EOI to clear ISR.
1568  *
1569  * If there are pending IRR bits they turn into ISR bits after a higher
1570  * priority ISR bit has been acked.
1571  */
1572 static void apic_pending_intr_clear(void)
1573 {
1574 	union apic_ir irr, isr;
1575 	unsigned int i;
1576 
1577 	/* 512 loops are way oversized and give the APIC a chance to obey. */
1578 	for (i = 0; i < 512; i++) {
1579 		if (!apic_check_and_ack(&irr, &isr))
1580 			return;
1581 	}
1582 	/* Dump the IRR/ISR content if that failed */
1583 	pr_warn("APIC: Stale IRR: %256pb ISR: %256pb\n", irr.map, isr.map);
1584 }
1585 
1586 /**
1587  * setup_local_APIC - setup the local APIC
1588  *
1589  * Used to setup local APIC while initializing BSP or bringing up APs.
1590  * Always called with preemption disabled.
1591  */
1592 static void setup_local_APIC(void)
1593 {
1594 	int cpu = smp_processor_id();
1595 	unsigned int value;
1596 
1597 	if (disable_apic) {
1598 		disable_ioapic_support();
1599 		return;
1600 	}
1601 
1602 	/*
1603 	 * If this comes from kexec/kcrash the APIC might be enabled in
1604 	 * SPIV. Soft disable it before doing further initialization.
1605 	 */
1606 	value = apic_read(APIC_SPIV);
1607 	value &= ~APIC_SPIV_APIC_ENABLED;
1608 	apic_write(APIC_SPIV, value);
1609 
1610 #ifdef CONFIG_X86_32
1611 	/* Pound the ESR really hard over the head with a big hammer - mbligh */
1612 	if (lapic_is_integrated() && apic->disable_esr) {
1613 		apic_write(APIC_ESR, 0);
1614 		apic_write(APIC_ESR, 0);
1615 		apic_write(APIC_ESR, 0);
1616 		apic_write(APIC_ESR, 0);
1617 	}
1618 #endif
1619 	/*
1620 	 * Double-check whether this APIC is really registered.
1621 	 * This is meaningless in clustered apic mode, so we skip it.
1622 	 */
1623 	BUG_ON(!apic->apic_id_registered());
1624 
1625 	/*
1626 	 * Intel recommends to set DFR, LDR and TPR before enabling
1627 	 * an APIC.  See e.g. "AP-388 82489DX User's Manual" (Intel
1628 	 * document number 292116).  So here it goes...
1629 	 */
1630 	apic->init_apic_ldr();
1631 
1632 #ifdef CONFIG_X86_32
1633 	if (apic->dest_logical) {
1634 		int logical_apicid, ldr_apicid;
1635 
1636 		/*
1637 		 * APIC LDR is initialized.  If logical_apicid mapping was
1638 		 * initialized during get_smp_config(), make sure it matches
1639 		 * the actual value.
1640 		 */
1641 		logical_apicid = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
1642 		ldr_apicid = GET_APIC_LOGICAL_ID(apic_read(APIC_LDR));
1643 		if (logical_apicid != BAD_APICID)
1644 			WARN_ON(logical_apicid != ldr_apicid);
1645 		/* Always use the value from LDR. */
1646 		early_per_cpu(x86_cpu_to_logical_apicid, cpu) = ldr_apicid;
1647 	}
1648 #endif
1649 
1650 	/*
1651 	 * Set Task Priority to 'accept all except vectors 0-31'.  An APIC
1652 	 * vector in the 16-31 range could be delivered if TPR == 0, but we
1653 	 * would think it's an exception and terrible things will happen.  We
1654 	 * never change this later on.
1655 	 */
1656 	value = apic_read(APIC_TASKPRI);
1657 	value &= ~APIC_TPRI_MASK;
1658 	value |= 0x10;
1659 	apic_write(APIC_TASKPRI, value);
1660 
1661 	/* Clear eventually stale ISR/IRR bits */
1662 	apic_pending_intr_clear();
1663 
1664 	/*
1665 	 * Now that we are all set up, enable the APIC
1666 	 */
1667 	value = apic_read(APIC_SPIV);
1668 	value &= ~APIC_VECTOR_MASK;
1669 	/*
1670 	 * Enable APIC
1671 	 */
1672 	value |= APIC_SPIV_APIC_ENABLED;
1673 
1674 #ifdef CONFIG_X86_32
1675 	/*
1676 	 * Some unknown Intel IO/APIC (or APIC) errata is biting us with
1677 	 * certain networking cards. If high frequency interrupts are
1678 	 * happening on a particular IOAPIC pin, plus the IOAPIC routing
1679 	 * entry is masked/unmasked at a high rate as well then sooner or
1680 	 * later IOAPIC line gets 'stuck', no more interrupts are received
1681 	 * from the device. If focus CPU is disabled then the hang goes
1682 	 * away, oh well :-(
1683 	 *
1684 	 * [ This bug can be reproduced easily with a level-triggered
1685 	 *   PCI Ne2000 networking cards and PII/PIII processors, dual
1686 	 *   BX chipset. ]
1687 	 */
1688 	/*
1689 	 * Actually disabling the focus CPU check just makes the hang less
1690 	 * frequent as it makes the interrupt distributon model be more
1691 	 * like LRU than MRU (the short-term load is more even across CPUs).
1692 	 */
1693 
1694 	/*
1695 	 * - enable focus processor (bit==0)
1696 	 * - 64bit mode always use processor focus
1697 	 *   so no need to set it
1698 	 */
1699 	value &= ~APIC_SPIV_FOCUS_DISABLED;
1700 #endif
1701 
1702 	/*
1703 	 * Set spurious IRQ vector
1704 	 */
1705 	value |= SPURIOUS_APIC_VECTOR;
1706 	apic_write(APIC_SPIV, value);
1707 
1708 	perf_events_lapic_init();
1709 
1710 	/*
1711 	 * Set up LVT0, LVT1:
1712 	 *
1713 	 * set up through-local-APIC on the boot CPU's LINT0. This is not
1714 	 * strictly necessary in pure symmetric-IO mode, but sometimes
1715 	 * we delegate interrupts to the 8259A.
1716 	 */
1717 	/*
1718 	 * TODO: set up through-local-APIC from through-I/O-APIC? --macro
1719 	 */
1720 	value = apic_read(APIC_LVT0) & APIC_LVT_MASKED;
1721 	if (!cpu && (pic_mode || !value || skip_ioapic_setup)) {
1722 		value = APIC_DM_EXTINT;
1723 		apic_printk(APIC_VERBOSE, "enabled ExtINT on CPU#%d\n", cpu);
1724 	} else {
1725 		value = APIC_DM_EXTINT | APIC_LVT_MASKED;
1726 		apic_printk(APIC_VERBOSE, "masked ExtINT on CPU#%d\n", cpu);
1727 	}
1728 	apic_write(APIC_LVT0, value);
1729 
1730 	/*
1731 	 * Only the BSP sees the LINT1 NMI signal by default. This can be
1732 	 * modified by apic_extnmi= boot option.
1733 	 */
1734 	if ((!cpu && apic_extnmi != APIC_EXTNMI_NONE) ||
1735 	    apic_extnmi == APIC_EXTNMI_ALL)
1736 		value = APIC_DM_NMI;
1737 	else
1738 		value = APIC_DM_NMI | APIC_LVT_MASKED;
1739 
1740 	/* Is 82489DX ? */
1741 	if (!lapic_is_integrated())
1742 		value |= APIC_LVT_LEVEL_TRIGGER;
1743 	apic_write(APIC_LVT1, value);
1744 
1745 #ifdef CONFIG_X86_MCE_INTEL
1746 	/* Recheck CMCI information after local APIC is up on CPU #0 */
1747 	if (!cpu)
1748 		cmci_recheck();
1749 #endif
1750 }
1751 
1752 static void end_local_APIC_setup(void)
1753 {
1754 	lapic_setup_esr();
1755 
1756 #ifdef CONFIG_X86_32
1757 	{
1758 		unsigned int value;
1759 		/* Disable the local apic timer */
1760 		value = apic_read(APIC_LVTT);
1761 		value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
1762 		apic_write(APIC_LVTT, value);
1763 	}
1764 #endif
1765 
1766 	apic_pm_activate();
1767 }
1768 
1769 /*
1770  * APIC setup function for application processors. Called from smpboot.c
1771  */
1772 void apic_ap_setup(void)
1773 {
1774 	setup_local_APIC();
1775 	end_local_APIC_setup();
1776 }
1777 
1778 #ifdef CONFIG_X86_X2APIC
1779 int x2apic_mode;
1780 
1781 enum {
1782 	X2APIC_OFF,
1783 	X2APIC_ON,
1784 	X2APIC_DISABLED,
1785 };
1786 static int x2apic_state;
1787 
1788 static void __x2apic_disable(void)
1789 {
1790 	u64 msr;
1791 
1792 	if (!boot_cpu_has(X86_FEATURE_APIC))
1793 		return;
1794 
1795 	rdmsrl(MSR_IA32_APICBASE, msr);
1796 	if (!(msr & X2APIC_ENABLE))
1797 		return;
1798 	/* Disable xapic and x2apic first and then reenable xapic mode */
1799 	wrmsrl(MSR_IA32_APICBASE, msr & ~(X2APIC_ENABLE | XAPIC_ENABLE));
1800 	wrmsrl(MSR_IA32_APICBASE, msr & ~X2APIC_ENABLE);
1801 	printk_once(KERN_INFO "x2apic disabled\n");
1802 }
1803 
1804 static void __x2apic_enable(void)
1805 {
1806 	u64 msr;
1807 
1808 	rdmsrl(MSR_IA32_APICBASE, msr);
1809 	if (msr & X2APIC_ENABLE)
1810 		return;
1811 	wrmsrl(MSR_IA32_APICBASE, msr | X2APIC_ENABLE);
1812 	printk_once(KERN_INFO "x2apic enabled\n");
1813 }
1814 
1815 static int __init setup_nox2apic(char *str)
1816 {
1817 	if (x2apic_enabled()) {
1818 		int apicid = native_apic_msr_read(APIC_ID);
1819 
1820 		if (apicid >= 255) {
1821 			pr_warn("Apicid: %08x, cannot enforce nox2apic\n",
1822 				apicid);
1823 			return 0;
1824 		}
1825 		pr_warn("x2apic already enabled.\n");
1826 		__x2apic_disable();
1827 	}
1828 	setup_clear_cpu_cap(X86_FEATURE_X2APIC);
1829 	x2apic_state = X2APIC_DISABLED;
1830 	x2apic_mode = 0;
1831 	return 0;
1832 }
1833 early_param("nox2apic", setup_nox2apic);
1834 
1835 /* Called from cpu_init() to enable x2apic on (secondary) cpus */
1836 void x2apic_setup(void)
1837 {
1838 	/*
1839 	 * If x2apic is not in ON state, disable it if already enabled
1840 	 * from BIOS.
1841 	 */
1842 	if (x2apic_state != X2APIC_ON) {
1843 		__x2apic_disable();
1844 		return;
1845 	}
1846 	__x2apic_enable();
1847 }
1848 
1849 static __init void x2apic_disable(void)
1850 {
1851 	u32 x2apic_id, state = x2apic_state;
1852 
1853 	x2apic_mode = 0;
1854 	x2apic_state = X2APIC_DISABLED;
1855 
1856 	if (state != X2APIC_ON)
1857 		return;
1858 
1859 	x2apic_id = read_apic_id();
1860 	if (x2apic_id >= 255)
1861 		panic("Cannot disable x2apic, id: %08x\n", x2apic_id);
1862 
1863 	__x2apic_disable();
1864 	register_lapic_address(mp_lapic_addr);
1865 }
1866 
1867 static __init void x2apic_enable(void)
1868 {
1869 	if (x2apic_state != X2APIC_OFF)
1870 		return;
1871 
1872 	x2apic_mode = 1;
1873 	x2apic_state = X2APIC_ON;
1874 	__x2apic_enable();
1875 }
1876 
1877 static __init void try_to_enable_x2apic(int remap_mode)
1878 {
1879 	if (x2apic_state == X2APIC_DISABLED)
1880 		return;
1881 
1882 	if (remap_mode != IRQ_REMAP_X2APIC_MODE) {
1883 		/* IR is required if there is APIC ID > 255 even when running
1884 		 * under KVM
1885 		 */
1886 		if (max_physical_apicid > 255 ||
1887 		    !x86_init.hyper.x2apic_available()) {
1888 			pr_info("x2apic: IRQ remapping doesn't support X2APIC mode\n");
1889 			x2apic_disable();
1890 			return;
1891 		}
1892 
1893 		/*
1894 		 * without IR all CPUs can be addressed by IOAPIC/MSI
1895 		 * only in physical mode
1896 		 */
1897 		x2apic_phys = 1;
1898 	}
1899 	x2apic_enable();
1900 }
1901 
1902 void __init check_x2apic(void)
1903 {
1904 	if (x2apic_enabled()) {
1905 		pr_info("x2apic: enabled by BIOS, switching to x2apic ops\n");
1906 		x2apic_mode = 1;
1907 		x2apic_state = X2APIC_ON;
1908 	} else if (!boot_cpu_has(X86_FEATURE_X2APIC)) {
1909 		x2apic_state = X2APIC_DISABLED;
1910 	}
1911 }
1912 #else /* CONFIG_X86_X2APIC */
1913 static int __init validate_x2apic(void)
1914 {
1915 	if (!apic_is_x2apic_enabled())
1916 		return 0;
1917 	/*
1918 	 * Checkme: Can we simply turn off x2apic here instead of panic?
1919 	 */
1920 	panic("BIOS has enabled x2apic but kernel doesn't support x2apic, please disable x2apic in BIOS.\n");
1921 }
1922 early_initcall(validate_x2apic);
1923 
1924 static inline void try_to_enable_x2apic(int remap_mode) { }
1925 static inline void __x2apic_enable(void) { }
1926 #endif /* !CONFIG_X86_X2APIC */
1927 
1928 void __init enable_IR_x2apic(void)
1929 {
1930 	unsigned long flags;
1931 	int ret, ir_stat;
1932 
1933 	if (skip_ioapic_setup) {
1934 		pr_info("Not enabling interrupt remapping due to skipped IO-APIC setup\n");
1935 		return;
1936 	}
1937 
1938 	ir_stat = irq_remapping_prepare();
1939 	if (ir_stat < 0 && !x2apic_supported())
1940 		return;
1941 
1942 	ret = save_ioapic_entries();
1943 	if (ret) {
1944 		pr_info("Saving IO-APIC state failed: %d\n", ret);
1945 		return;
1946 	}
1947 
1948 	local_irq_save(flags);
1949 	legacy_pic->mask_all();
1950 	mask_ioapic_entries();
1951 
1952 	/* If irq_remapping_prepare() succeeded, try to enable it */
1953 	if (ir_stat >= 0)
1954 		ir_stat = irq_remapping_enable();
1955 	/* ir_stat contains the remap mode or an error code */
1956 	try_to_enable_x2apic(ir_stat);
1957 
1958 	if (ir_stat < 0)
1959 		restore_ioapic_entries();
1960 	legacy_pic->restore_mask();
1961 	local_irq_restore(flags);
1962 }
1963 
1964 #ifdef CONFIG_X86_64
1965 /*
1966  * Detect and enable local APICs on non-SMP boards.
1967  * Original code written by Keir Fraser.
1968  * On AMD64 we trust the BIOS - if it says no APIC it is likely
1969  * not correctly set up (usually the APIC timer won't work etc.)
1970  */
1971 static int __init detect_init_APIC(void)
1972 {
1973 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1974 		pr_info("No local APIC present\n");
1975 		return -1;
1976 	}
1977 
1978 	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1979 	return 0;
1980 }
1981 #else
1982 
1983 static int __init apic_verify(void)
1984 {
1985 	u32 features, h, l;
1986 
1987 	/*
1988 	 * The APIC feature bit should now be enabled
1989 	 * in `cpuid'
1990 	 */
1991 	features = cpuid_edx(1);
1992 	if (!(features & (1 << X86_FEATURE_APIC))) {
1993 		pr_warn("Could not enable APIC!\n");
1994 		return -1;
1995 	}
1996 	set_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
1997 	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1998 
1999 	/* The BIOS may have set up the APIC at some other address */
2000 	if (boot_cpu_data.x86 >= 6) {
2001 		rdmsr(MSR_IA32_APICBASE, l, h);
2002 		if (l & MSR_IA32_APICBASE_ENABLE)
2003 			mp_lapic_addr = l & MSR_IA32_APICBASE_BASE;
2004 	}
2005 
2006 	pr_info("Found and enabled local APIC!\n");
2007 	return 0;
2008 }
2009 
2010 int __init apic_force_enable(unsigned long addr)
2011 {
2012 	u32 h, l;
2013 
2014 	if (disable_apic)
2015 		return -1;
2016 
2017 	/*
2018 	 * Some BIOSes disable the local APIC in the APIC_BASE
2019 	 * MSR. This can only be done in software for Intel P6 or later
2020 	 * and AMD K7 (Model > 1) or later.
2021 	 */
2022 	if (boot_cpu_data.x86 >= 6) {
2023 		rdmsr(MSR_IA32_APICBASE, l, h);
2024 		if (!(l & MSR_IA32_APICBASE_ENABLE)) {
2025 			pr_info("Local APIC disabled by BIOS -- reenabling.\n");
2026 			l &= ~MSR_IA32_APICBASE_BASE;
2027 			l |= MSR_IA32_APICBASE_ENABLE | addr;
2028 			wrmsr(MSR_IA32_APICBASE, l, h);
2029 			enabled_via_apicbase = 1;
2030 		}
2031 	}
2032 	return apic_verify();
2033 }
2034 
2035 /*
2036  * Detect and initialize APIC
2037  */
2038 static int __init detect_init_APIC(void)
2039 {
2040 	/* Disabled by kernel option? */
2041 	if (disable_apic)
2042 		return -1;
2043 
2044 	switch (boot_cpu_data.x86_vendor) {
2045 	case X86_VENDOR_AMD:
2046 		if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model > 1) ||
2047 		    (boot_cpu_data.x86 >= 15))
2048 			break;
2049 		goto no_apic;
2050 	case X86_VENDOR_HYGON:
2051 		break;
2052 	case X86_VENDOR_INTEL:
2053 		if (boot_cpu_data.x86 == 6 || boot_cpu_data.x86 == 15 ||
2054 		    (boot_cpu_data.x86 == 5 && boot_cpu_has(X86_FEATURE_APIC)))
2055 			break;
2056 		goto no_apic;
2057 	default:
2058 		goto no_apic;
2059 	}
2060 
2061 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
2062 		/*
2063 		 * Over-ride BIOS and try to enable the local APIC only if
2064 		 * "lapic" specified.
2065 		 */
2066 		if (!force_enable_local_apic) {
2067 			pr_info("Local APIC disabled by BIOS -- "
2068 				"you can enable it with \"lapic\"\n");
2069 			return -1;
2070 		}
2071 		if (apic_force_enable(APIC_DEFAULT_PHYS_BASE))
2072 			return -1;
2073 	} else {
2074 		if (apic_verify())
2075 			return -1;
2076 	}
2077 
2078 	apic_pm_activate();
2079 
2080 	return 0;
2081 
2082 no_apic:
2083 	pr_info("No local APIC present or hardware disabled\n");
2084 	return -1;
2085 }
2086 #endif
2087 
2088 /**
2089  * init_apic_mappings - initialize APIC mappings
2090  */
2091 void __init init_apic_mappings(void)
2092 {
2093 	unsigned int new_apicid;
2094 
2095 	if (apic_validate_deadline_timer())
2096 		pr_debug("TSC deadline timer available\n");
2097 
2098 	if (x2apic_mode) {
2099 		boot_cpu_physical_apicid = read_apic_id();
2100 		return;
2101 	}
2102 
2103 	/* If no local APIC can be found return early */
2104 	if (!smp_found_config && detect_init_APIC()) {
2105 		/* lets NOP'ify apic operations */
2106 		pr_info("APIC: disable apic facility\n");
2107 		apic_disable();
2108 	} else {
2109 		apic_phys = mp_lapic_addr;
2110 
2111 		/*
2112 		 * If the system has ACPI MADT tables or MP info, the LAPIC
2113 		 * address is already registered.
2114 		 */
2115 		if (!acpi_lapic && !smp_found_config)
2116 			register_lapic_address(apic_phys);
2117 	}
2118 
2119 	/*
2120 	 * Fetch the APIC ID of the BSP in case we have a
2121 	 * default configuration (or the MP table is broken).
2122 	 */
2123 	new_apicid = read_apic_id();
2124 	if (boot_cpu_physical_apicid != new_apicid) {
2125 		boot_cpu_physical_apicid = new_apicid;
2126 		/*
2127 		 * yeah -- we lie about apic_version
2128 		 * in case if apic was disabled via boot option
2129 		 * but it's not a problem for SMP compiled kernel
2130 		 * since apic_intr_mode_select is prepared for such
2131 		 * a case and disable smp mode
2132 		 */
2133 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
2134 	}
2135 }
2136 
2137 void __init register_lapic_address(unsigned long address)
2138 {
2139 	mp_lapic_addr = address;
2140 
2141 	if (!x2apic_mode) {
2142 		set_fixmap_nocache(FIX_APIC_BASE, address);
2143 		apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
2144 			    APIC_BASE, address);
2145 	}
2146 	if (boot_cpu_physical_apicid == -1U) {
2147 		boot_cpu_physical_apicid  = read_apic_id();
2148 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
2149 	}
2150 }
2151 
2152 /*
2153  * Local APIC interrupts
2154  */
2155 
2156 /*
2157  * This interrupt should _never_ happen with our APIC/SMP architecture
2158  */
2159 __visible void __irq_entry smp_spurious_interrupt(struct pt_regs *regs)
2160 {
2161 	u8 vector = ~regs->orig_ax;
2162 	u32 v;
2163 
2164 	entering_irq();
2165 	trace_spurious_apic_entry(vector);
2166 
2167 	inc_irq_stat(irq_spurious_count);
2168 
2169 	/*
2170 	 * If this is a spurious interrupt then do not acknowledge
2171 	 */
2172 	if (vector == SPURIOUS_APIC_VECTOR) {
2173 		/* See SDM vol 3 */
2174 		pr_info("Spurious APIC interrupt (vector 0xFF) on CPU#%d, should never happen.\n",
2175 			smp_processor_id());
2176 		goto out;
2177 	}
2178 
2179 	/*
2180 	 * If it is a vectored one, verify it's set in the ISR. If set,
2181 	 * acknowledge it.
2182 	 */
2183 	v = apic_read(APIC_ISR + ((vector & ~0x1f) >> 1));
2184 	if (v & (1 << (vector & 0x1f))) {
2185 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Acked\n",
2186 			vector, smp_processor_id());
2187 		ack_APIC_irq();
2188 	} else {
2189 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Not pending!\n",
2190 			vector, smp_processor_id());
2191 	}
2192 out:
2193 	trace_spurious_apic_exit(vector);
2194 	exiting_irq();
2195 }
2196 
2197 /*
2198  * This interrupt should never happen with our APIC/SMP architecture
2199  */
2200 __visible void __irq_entry smp_error_interrupt(struct pt_regs *regs)
2201 {
2202 	static const char * const error_interrupt_reason[] = {
2203 		"Send CS error",		/* APIC Error Bit 0 */
2204 		"Receive CS error",		/* APIC Error Bit 1 */
2205 		"Send accept error",		/* APIC Error Bit 2 */
2206 		"Receive accept error",		/* APIC Error Bit 3 */
2207 		"Redirectable IPI",		/* APIC Error Bit 4 */
2208 		"Send illegal vector",		/* APIC Error Bit 5 */
2209 		"Received illegal vector",	/* APIC Error Bit 6 */
2210 		"Illegal register address",	/* APIC Error Bit 7 */
2211 	};
2212 	u32 v, i = 0;
2213 
2214 	entering_irq();
2215 	trace_error_apic_entry(ERROR_APIC_VECTOR);
2216 
2217 	/* First tickle the hardware, only then report what went on. -- REW */
2218 	if (lapic_get_maxlvt() > 3)	/* Due to the Pentium erratum 3AP. */
2219 		apic_write(APIC_ESR, 0);
2220 	v = apic_read(APIC_ESR);
2221 	ack_APIC_irq();
2222 	atomic_inc(&irq_err_count);
2223 
2224 	apic_printk(APIC_DEBUG, KERN_DEBUG "APIC error on CPU%d: %02x",
2225 		    smp_processor_id(), v);
2226 
2227 	v &= 0xff;
2228 	while (v) {
2229 		if (v & 0x1)
2230 			apic_printk(APIC_DEBUG, KERN_CONT " : %s", error_interrupt_reason[i]);
2231 		i++;
2232 		v >>= 1;
2233 	}
2234 
2235 	apic_printk(APIC_DEBUG, KERN_CONT "\n");
2236 
2237 	trace_error_apic_exit(ERROR_APIC_VECTOR);
2238 	exiting_irq();
2239 }
2240 
2241 /**
2242  * connect_bsp_APIC - attach the APIC to the interrupt system
2243  */
2244 static void __init connect_bsp_APIC(void)
2245 {
2246 #ifdef CONFIG_X86_32
2247 	if (pic_mode) {
2248 		/*
2249 		 * Do not trust the local APIC being empty at bootup.
2250 		 */
2251 		clear_local_APIC();
2252 		/*
2253 		 * PIC mode, enable APIC mode in the IMCR, i.e.  connect BSP's
2254 		 * local APIC to INT and NMI lines.
2255 		 */
2256 		apic_printk(APIC_VERBOSE, "leaving PIC mode, "
2257 				"enabling APIC mode.\n");
2258 		imcr_pic_to_apic();
2259 	}
2260 #endif
2261 }
2262 
2263 /**
2264  * disconnect_bsp_APIC - detach the APIC from the interrupt system
2265  * @virt_wire_setup:	indicates, whether virtual wire mode is selected
2266  *
2267  * Virtual wire mode is necessary to deliver legacy interrupts even when the
2268  * APIC is disabled.
2269  */
2270 void disconnect_bsp_APIC(int virt_wire_setup)
2271 {
2272 	unsigned int value;
2273 
2274 #ifdef CONFIG_X86_32
2275 	if (pic_mode) {
2276 		/*
2277 		 * Put the board back into PIC mode (has an effect only on
2278 		 * certain older boards).  Note that APIC interrupts, including
2279 		 * IPIs, won't work beyond this point!  The only exception are
2280 		 * INIT IPIs.
2281 		 */
2282 		apic_printk(APIC_VERBOSE, "disabling APIC mode, "
2283 				"entering PIC mode.\n");
2284 		imcr_apic_to_pic();
2285 		return;
2286 	}
2287 #endif
2288 
2289 	/* Go back to Virtual Wire compatibility mode */
2290 
2291 	/* For the spurious interrupt use vector F, and enable it */
2292 	value = apic_read(APIC_SPIV);
2293 	value &= ~APIC_VECTOR_MASK;
2294 	value |= APIC_SPIV_APIC_ENABLED;
2295 	value |= 0xf;
2296 	apic_write(APIC_SPIV, value);
2297 
2298 	if (!virt_wire_setup) {
2299 		/*
2300 		 * For LVT0 make it edge triggered, active high,
2301 		 * external and enabled
2302 		 */
2303 		value = apic_read(APIC_LVT0);
2304 		value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2305 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2306 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2307 		value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2308 		value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_EXTINT);
2309 		apic_write(APIC_LVT0, value);
2310 	} else {
2311 		/* Disable LVT0 */
2312 		apic_write(APIC_LVT0, APIC_LVT_MASKED);
2313 	}
2314 
2315 	/*
2316 	 * For LVT1 make it edge triggered, active high,
2317 	 * nmi and enabled
2318 	 */
2319 	value = apic_read(APIC_LVT1);
2320 	value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2321 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2322 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2323 	value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2324 	value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_NMI);
2325 	apic_write(APIC_LVT1, value);
2326 }
2327 
2328 /*
2329  * The number of allocated logical CPU IDs. Since logical CPU IDs are allocated
2330  * contiguously, it equals to current allocated max logical CPU ID plus 1.
2331  * All allocated CPU IDs should be in the [0, nr_logical_cpuids) range,
2332  * so the maximum of nr_logical_cpuids is nr_cpu_ids.
2333  *
2334  * NOTE: Reserve 0 for BSP.
2335  */
2336 static int nr_logical_cpuids = 1;
2337 
2338 /*
2339  * Used to store mapping between logical CPU IDs and APIC IDs.
2340  */
2341 static int cpuid_to_apicid[] = {
2342 	[0 ... NR_CPUS - 1] = -1,
2343 };
2344 
2345 #ifdef CONFIG_SMP
2346 /**
2347  * apic_id_is_primary_thread - Check whether APIC ID belongs to a primary thread
2348  * @apicid: APIC ID to check
2349  */
2350 bool apic_id_is_primary_thread(unsigned int apicid)
2351 {
2352 	u32 mask;
2353 
2354 	if (smp_num_siblings == 1)
2355 		return true;
2356 	/* Isolate the SMT bit(s) in the APICID and check for 0 */
2357 	mask = (1U << (fls(smp_num_siblings) - 1)) - 1;
2358 	return !(apicid & mask);
2359 }
2360 #endif
2361 
2362 /*
2363  * Should use this API to allocate logical CPU IDs to keep nr_logical_cpuids
2364  * and cpuid_to_apicid[] synchronized.
2365  */
2366 static int allocate_logical_cpuid(int apicid)
2367 {
2368 	int i;
2369 
2370 	/*
2371 	 * cpuid <-> apicid mapping is persistent, so when a cpu is up,
2372 	 * check if the kernel has allocated a cpuid for it.
2373 	 */
2374 	for (i = 0; i < nr_logical_cpuids; i++) {
2375 		if (cpuid_to_apicid[i] == apicid)
2376 			return i;
2377 	}
2378 
2379 	/* Allocate a new cpuid. */
2380 	if (nr_logical_cpuids >= nr_cpu_ids) {
2381 		WARN_ONCE(1, "APIC: NR_CPUS/possible_cpus limit of %u reached. "
2382 			     "Processor %d/0x%x and the rest are ignored.\n",
2383 			     nr_cpu_ids, nr_logical_cpuids, apicid);
2384 		return -EINVAL;
2385 	}
2386 
2387 	cpuid_to_apicid[nr_logical_cpuids] = apicid;
2388 	return nr_logical_cpuids++;
2389 }
2390 
2391 int generic_processor_info(int apicid, int version)
2392 {
2393 	int cpu, max = nr_cpu_ids;
2394 	bool boot_cpu_detected = physid_isset(boot_cpu_physical_apicid,
2395 				phys_cpu_present_map);
2396 
2397 	/*
2398 	 * boot_cpu_physical_apicid is designed to have the apicid
2399 	 * returned by read_apic_id(), i.e, the apicid of the
2400 	 * currently booting-up processor. However, on some platforms,
2401 	 * it is temporarily modified by the apicid reported as BSP
2402 	 * through MP table. Concretely:
2403 	 *
2404 	 * - arch/x86/kernel/mpparse.c: MP_processor_info()
2405 	 * - arch/x86/mm/amdtopology.c: amd_numa_init()
2406 	 *
2407 	 * This function is executed with the modified
2408 	 * boot_cpu_physical_apicid. So, disabled_cpu_apicid kernel
2409 	 * parameter doesn't work to disable APs on kdump 2nd kernel.
2410 	 *
2411 	 * Since fixing handling of boot_cpu_physical_apicid requires
2412 	 * another discussion and tests on each platform, we leave it
2413 	 * for now and here we use read_apic_id() directly in this
2414 	 * function, generic_processor_info().
2415 	 */
2416 	if (disabled_cpu_apicid != BAD_APICID &&
2417 	    disabled_cpu_apicid != read_apic_id() &&
2418 	    disabled_cpu_apicid == apicid) {
2419 		int thiscpu = num_processors + disabled_cpus;
2420 
2421 		pr_warn("APIC: Disabling requested cpu."
2422 			" Processor %d/0x%x ignored.\n", thiscpu, apicid);
2423 
2424 		disabled_cpus++;
2425 		return -ENODEV;
2426 	}
2427 
2428 	/*
2429 	 * If boot cpu has not been detected yet, then only allow upto
2430 	 * nr_cpu_ids - 1 processors and keep one slot free for boot cpu
2431 	 */
2432 	if (!boot_cpu_detected && num_processors >= nr_cpu_ids - 1 &&
2433 	    apicid != boot_cpu_physical_apicid) {
2434 		int thiscpu = max + disabled_cpus - 1;
2435 
2436 		pr_warn("APIC: NR_CPUS/possible_cpus limit of %i almost"
2437 			" reached. Keeping one slot for boot cpu."
2438 			"  Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
2439 
2440 		disabled_cpus++;
2441 		return -ENODEV;
2442 	}
2443 
2444 	if (num_processors >= nr_cpu_ids) {
2445 		int thiscpu = max + disabled_cpus;
2446 
2447 		pr_warn("APIC: NR_CPUS/possible_cpus limit of %i reached. "
2448 			"Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
2449 
2450 		disabled_cpus++;
2451 		return -EINVAL;
2452 	}
2453 
2454 	if (apicid == boot_cpu_physical_apicid) {
2455 		/*
2456 		 * x86_bios_cpu_apicid is required to have processors listed
2457 		 * in same order as logical cpu numbers. Hence the first
2458 		 * entry is BSP, and so on.
2459 		 * boot_cpu_init() already hold bit 0 in cpu_present_mask
2460 		 * for BSP.
2461 		 */
2462 		cpu = 0;
2463 
2464 		/* Logical cpuid 0 is reserved for BSP. */
2465 		cpuid_to_apicid[0] = apicid;
2466 	} else {
2467 		cpu = allocate_logical_cpuid(apicid);
2468 		if (cpu < 0) {
2469 			disabled_cpus++;
2470 			return -EINVAL;
2471 		}
2472 	}
2473 
2474 	/*
2475 	 * Validate version
2476 	 */
2477 	if (version == 0x0) {
2478 		pr_warn("BIOS bug: APIC version is 0 for CPU %d/0x%x, fixing up to 0x10\n",
2479 			cpu, apicid);
2480 		version = 0x10;
2481 	}
2482 
2483 	if (version != boot_cpu_apic_version) {
2484 		pr_warn("BIOS bug: APIC version mismatch, boot CPU: %x, CPU %d: version %x\n",
2485 			boot_cpu_apic_version, cpu, version);
2486 	}
2487 
2488 	if (apicid > max_physical_apicid)
2489 		max_physical_apicid = apicid;
2490 
2491 #if defined(CONFIG_SMP) || defined(CONFIG_X86_64)
2492 	early_per_cpu(x86_cpu_to_apicid, cpu) = apicid;
2493 	early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
2494 #endif
2495 #ifdef CONFIG_X86_32
2496 	early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
2497 		apic->x86_32_early_logical_apicid(cpu);
2498 #endif
2499 	set_cpu_possible(cpu, true);
2500 	physid_set(apicid, phys_cpu_present_map);
2501 	set_cpu_present(cpu, true);
2502 	num_processors++;
2503 
2504 	return cpu;
2505 }
2506 
2507 int hard_smp_processor_id(void)
2508 {
2509 	return read_apic_id();
2510 }
2511 
2512 /*
2513  * Override the generic EOI implementation with an optimized version.
2514  * Only called during early boot when only one CPU is active and with
2515  * interrupts disabled, so we know this does not race with actual APIC driver
2516  * use.
2517  */
2518 void __init apic_set_eoi_write(void (*eoi_write)(u32 reg, u32 v))
2519 {
2520 	struct apic **drv;
2521 
2522 	for (drv = __apicdrivers; drv < __apicdrivers_end; drv++) {
2523 		/* Should happen once for each apic */
2524 		WARN_ON((*drv)->eoi_write == eoi_write);
2525 		(*drv)->native_eoi_write = (*drv)->eoi_write;
2526 		(*drv)->eoi_write = eoi_write;
2527 	}
2528 }
2529 
2530 static void __init apic_bsp_up_setup(void)
2531 {
2532 #ifdef CONFIG_X86_64
2533 	apic_write(APIC_ID, apic->set_apic_id(boot_cpu_physical_apicid));
2534 #else
2535 	/*
2536 	 * Hack: In case of kdump, after a crash, kernel might be booting
2537 	 * on a cpu with non-zero lapic id. But boot_cpu_physical_apicid
2538 	 * might be zero if read from MP tables. Get it from LAPIC.
2539 	 */
2540 # ifdef CONFIG_CRASH_DUMP
2541 	boot_cpu_physical_apicid = read_apic_id();
2542 # endif
2543 #endif
2544 	physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
2545 }
2546 
2547 /**
2548  * apic_bsp_setup - Setup function for local apic and io-apic
2549  * @upmode:		Force UP mode (for APIC_init_uniprocessor)
2550  */
2551 static void __init apic_bsp_setup(bool upmode)
2552 {
2553 	connect_bsp_APIC();
2554 	if (upmode)
2555 		apic_bsp_up_setup();
2556 	setup_local_APIC();
2557 
2558 	enable_IO_APIC();
2559 	end_local_APIC_setup();
2560 	irq_remap_enable_fault_handling();
2561 	setup_IO_APIC();
2562 }
2563 
2564 #ifdef CONFIG_UP_LATE_INIT
2565 void __init up_late_init(void)
2566 {
2567 	if (apic_intr_mode == APIC_PIC)
2568 		return;
2569 
2570 	/* Setup local timer */
2571 	x86_init.timers.setup_percpu_clockev();
2572 }
2573 #endif
2574 
2575 /*
2576  * Power management
2577  */
2578 #ifdef CONFIG_PM
2579 
2580 static struct {
2581 	/*
2582 	 * 'active' is true if the local APIC was enabled by us and
2583 	 * not the BIOS; this signifies that we are also responsible
2584 	 * for disabling it before entering apm/acpi suspend
2585 	 */
2586 	int active;
2587 	/* r/w apic fields */
2588 	unsigned int apic_id;
2589 	unsigned int apic_taskpri;
2590 	unsigned int apic_ldr;
2591 	unsigned int apic_dfr;
2592 	unsigned int apic_spiv;
2593 	unsigned int apic_lvtt;
2594 	unsigned int apic_lvtpc;
2595 	unsigned int apic_lvt0;
2596 	unsigned int apic_lvt1;
2597 	unsigned int apic_lvterr;
2598 	unsigned int apic_tmict;
2599 	unsigned int apic_tdcr;
2600 	unsigned int apic_thmr;
2601 	unsigned int apic_cmci;
2602 } apic_pm_state;
2603 
2604 static int lapic_suspend(void)
2605 {
2606 	unsigned long flags;
2607 	int maxlvt;
2608 
2609 	if (!apic_pm_state.active)
2610 		return 0;
2611 
2612 	maxlvt = lapic_get_maxlvt();
2613 
2614 	apic_pm_state.apic_id = apic_read(APIC_ID);
2615 	apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
2616 	apic_pm_state.apic_ldr = apic_read(APIC_LDR);
2617 	apic_pm_state.apic_dfr = apic_read(APIC_DFR);
2618 	apic_pm_state.apic_spiv = apic_read(APIC_SPIV);
2619 	apic_pm_state.apic_lvtt = apic_read(APIC_LVTT);
2620 	if (maxlvt >= 4)
2621 		apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC);
2622 	apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0);
2623 	apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1);
2624 	apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR);
2625 	apic_pm_state.apic_tmict = apic_read(APIC_TMICT);
2626 	apic_pm_state.apic_tdcr = apic_read(APIC_TDCR);
2627 #ifdef CONFIG_X86_THERMAL_VECTOR
2628 	if (maxlvt >= 5)
2629 		apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
2630 #endif
2631 #ifdef CONFIG_X86_MCE_INTEL
2632 	if (maxlvt >= 6)
2633 		apic_pm_state.apic_cmci = apic_read(APIC_LVTCMCI);
2634 #endif
2635 
2636 	local_irq_save(flags);
2637 
2638 	/*
2639 	 * Mask IOAPIC before disabling the local APIC to prevent stale IRR
2640 	 * entries on some implementations.
2641 	 */
2642 	mask_ioapic_entries();
2643 
2644 	disable_local_APIC();
2645 
2646 	irq_remapping_disable();
2647 
2648 	local_irq_restore(flags);
2649 	return 0;
2650 }
2651 
2652 static void lapic_resume(void)
2653 {
2654 	unsigned int l, h;
2655 	unsigned long flags;
2656 	int maxlvt;
2657 
2658 	if (!apic_pm_state.active)
2659 		return;
2660 
2661 	local_irq_save(flags);
2662 
2663 	/*
2664 	 * IO-APIC and PIC have their own resume routines.
2665 	 * We just mask them here to make sure the interrupt
2666 	 * subsystem is completely quiet while we enable x2apic
2667 	 * and interrupt-remapping.
2668 	 */
2669 	mask_ioapic_entries();
2670 	legacy_pic->mask_all();
2671 
2672 	if (x2apic_mode) {
2673 		__x2apic_enable();
2674 	} else {
2675 		/*
2676 		 * Make sure the APICBASE points to the right address
2677 		 *
2678 		 * FIXME! This will be wrong if we ever support suspend on
2679 		 * SMP! We'll need to do this as part of the CPU restore!
2680 		 */
2681 		if (boot_cpu_data.x86 >= 6) {
2682 			rdmsr(MSR_IA32_APICBASE, l, h);
2683 			l &= ~MSR_IA32_APICBASE_BASE;
2684 			l |= MSR_IA32_APICBASE_ENABLE | mp_lapic_addr;
2685 			wrmsr(MSR_IA32_APICBASE, l, h);
2686 		}
2687 	}
2688 
2689 	maxlvt = lapic_get_maxlvt();
2690 	apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED);
2691 	apic_write(APIC_ID, apic_pm_state.apic_id);
2692 	apic_write(APIC_DFR, apic_pm_state.apic_dfr);
2693 	apic_write(APIC_LDR, apic_pm_state.apic_ldr);
2694 	apic_write(APIC_TASKPRI, apic_pm_state.apic_taskpri);
2695 	apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
2696 	apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
2697 	apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
2698 #ifdef CONFIG_X86_THERMAL_VECTOR
2699 	if (maxlvt >= 5)
2700 		apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
2701 #endif
2702 #ifdef CONFIG_X86_MCE_INTEL
2703 	if (maxlvt >= 6)
2704 		apic_write(APIC_LVTCMCI, apic_pm_state.apic_cmci);
2705 #endif
2706 	if (maxlvt >= 4)
2707 		apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
2708 	apic_write(APIC_LVTT, apic_pm_state.apic_lvtt);
2709 	apic_write(APIC_TDCR, apic_pm_state.apic_tdcr);
2710 	apic_write(APIC_TMICT, apic_pm_state.apic_tmict);
2711 	apic_write(APIC_ESR, 0);
2712 	apic_read(APIC_ESR);
2713 	apic_write(APIC_LVTERR, apic_pm_state.apic_lvterr);
2714 	apic_write(APIC_ESR, 0);
2715 	apic_read(APIC_ESR);
2716 
2717 	irq_remapping_reenable(x2apic_mode);
2718 
2719 	local_irq_restore(flags);
2720 }
2721 
2722 /*
2723  * This device has no shutdown method - fully functioning local APICs
2724  * are needed on every CPU up until machine_halt/restart/poweroff.
2725  */
2726 
2727 static struct syscore_ops lapic_syscore_ops = {
2728 	.resume		= lapic_resume,
2729 	.suspend	= lapic_suspend,
2730 };
2731 
2732 static void apic_pm_activate(void)
2733 {
2734 	apic_pm_state.active = 1;
2735 }
2736 
2737 static int __init init_lapic_sysfs(void)
2738 {
2739 	/* XXX: remove suspend/resume procs if !apic_pm_state.active? */
2740 	if (boot_cpu_has(X86_FEATURE_APIC))
2741 		register_syscore_ops(&lapic_syscore_ops);
2742 
2743 	return 0;
2744 }
2745 
2746 /* local apic needs to resume before other devices access its registers. */
2747 core_initcall(init_lapic_sysfs);
2748 
2749 #else	/* CONFIG_PM */
2750 
2751 static void apic_pm_activate(void) { }
2752 
2753 #endif	/* CONFIG_PM */
2754 
2755 #ifdef CONFIG_X86_64
2756 
2757 static int multi_checked;
2758 static int multi;
2759 
2760 static int set_multi(const struct dmi_system_id *d)
2761 {
2762 	if (multi)
2763 		return 0;
2764 	pr_info("APIC: %s detected, Multi Chassis\n", d->ident);
2765 	multi = 1;
2766 	return 0;
2767 }
2768 
2769 static const struct dmi_system_id multi_dmi_table[] = {
2770 	{
2771 		.callback = set_multi,
2772 		.ident = "IBM System Summit2",
2773 		.matches = {
2774 			DMI_MATCH(DMI_SYS_VENDOR, "IBM"),
2775 			DMI_MATCH(DMI_PRODUCT_NAME, "Summit2"),
2776 		},
2777 	},
2778 	{}
2779 };
2780 
2781 static void dmi_check_multi(void)
2782 {
2783 	if (multi_checked)
2784 		return;
2785 
2786 	dmi_check_system(multi_dmi_table);
2787 	multi_checked = 1;
2788 }
2789 
2790 /*
2791  * apic_is_clustered_box() -- Check if we can expect good TSC
2792  *
2793  * Thus far, the major user of this is IBM's Summit2 series:
2794  * Clustered boxes may have unsynced TSC problems if they are
2795  * multi-chassis.
2796  * Use DMI to check them
2797  */
2798 int apic_is_clustered_box(void)
2799 {
2800 	dmi_check_multi();
2801 	return multi;
2802 }
2803 #endif
2804 
2805 /*
2806  * APIC command line parameters
2807  */
2808 static int __init setup_disableapic(char *arg)
2809 {
2810 	disable_apic = 1;
2811 	setup_clear_cpu_cap(X86_FEATURE_APIC);
2812 	return 0;
2813 }
2814 early_param("disableapic", setup_disableapic);
2815 
2816 /* same as disableapic, for compatibility */
2817 static int __init setup_nolapic(char *arg)
2818 {
2819 	return setup_disableapic(arg);
2820 }
2821 early_param("nolapic", setup_nolapic);
2822 
2823 static int __init parse_lapic_timer_c2_ok(char *arg)
2824 {
2825 	local_apic_timer_c2_ok = 1;
2826 	return 0;
2827 }
2828 early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok);
2829 
2830 static int __init parse_disable_apic_timer(char *arg)
2831 {
2832 	disable_apic_timer = 1;
2833 	return 0;
2834 }
2835 early_param("noapictimer", parse_disable_apic_timer);
2836 
2837 static int __init parse_nolapic_timer(char *arg)
2838 {
2839 	disable_apic_timer = 1;
2840 	return 0;
2841 }
2842 early_param("nolapic_timer", parse_nolapic_timer);
2843 
2844 static int __init apic_set_verbosity(char *arg)
2845 {
2846 	if (!arg)  {
2847 #ifdef CONFIG_X86_64
2848 		skip_ioapic_setup = 0;
2849 		return 0;
2850 #endif
2851 		return -EINVAL;
2852 	}
2853 
2854 	if (strcmp("debug", arg) == 0)
2855 		apic_verbosity = APIC_DEBUG;
2856 	else if (strcmp("verbose", arg) == 0)
2857 		apic_verbosity = APIC_VERBOSE;
2858 #ifdef CONFIG_X86_64
2859 	else {
2860 		pr_warn("APIC Verbosity level %s not recognised"
2861 			" use apic=verbose or apic=debug\n", arg);
2862 		return -EINVAL;
2863 	}
2864 #endif
2865 
2866 	return 0;
2867 }
2868 early_param("apic", apic_set_verbosity);
2869 
2870 static int __init lapic_insert_resource(void)
2871 {
2872 	if (!apic_phys)
2873 		return -1;
2874 
2875 	/* Put local APIC into the resource map. */
2876 	lapic_resource.start = apic_phys;
2877 	lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
2878 	insert_resource(&iomem_resource, &lapic_resource);
2879 
2880 	return 0;
2881 }
2882 
2883 /*
2884  * need call insert after e820__reserve_resources()
2885  * that is using request_resource
2886  */
2887 late_initcall(lapic_insert_resource);
2888 
2889 static int __init apic_set_disabled_cpu_apicid(char *arg)
2890 {
2891 	if (!arg || !get_option(&arg, &disabled_cpu_apicid))
2892 		return -EINVAL;
2893 
2894 	return 0;
2895 }
2896 early_param("disable_cpu_apicid", apic_set_disabled_cpu_apicid);
2897 
2898 static int __init apic_set_extnmi(char *arg)
2899 {
2900 	if (!arg)
2901 		return -EINVAL;
2902 
2903 	if (!strncmp("all", arg, 3))
2904 		apic_extnmi = APIC_EXTNMI_ALL;
2905 	else if (!strncmp("none", arg, 4))
2906 		apic_extnmi = APIC_EXTNMI_NONE;
2907 	else if (!strncmp("bsp", arg, 3))
2908 		apic_extnmi = APIC_EXTNMI_BSP;
2909 	else {
2910 		pr_warn("Unknown external NMI delivery mode `%s' ignored\n", arg);
2911 		return -EINVAL;
2912 	}
2913 
2914 	return 0;
2915 }
2916 early_param("apic_extnmi", apic_set_extnmi);
2917