xref: /openbmc/linux/arch/x86/kernel/apic/apic.c (revision 78560d41)
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 
356 		printk_once(KERN_DEBUG "TSC deadline timer enabled\n");
357 		return;
358 	}
359 
360 	/*
361 	 * Divide PICLK by 16
362 	 */
363 	tmp_value = apic_read(APIC_TDCR);
364 	apic_write(APIC_TDCR,
365 		(tmp_value & ~(APIC_TDR_DIV_1 | APIC_TDR_DIV_TMBASE)) |
366 		APIC_TDR_DIV_16);
367 
368 	if (!oneshot)
369 		apic_write(APIC_TMICT, clocks / APIC_DIVISOR);
370 }
371 
372 /*
373  * Setup extended LVT, AMD specific
374  *
375  * Software should use the LVT offsets the BIOS provides.  The offsets
376  * are determined by the subsystems using it like those for MCE
377  * threshold or IBS.  On K8 only offset 0 (APIC500) and MCE interrupts
378  * are supported. Beginning with family 10h at least 4 offsets are
379  * available.
380  *
381  * Since the offsets must be consistent for all cores, we keep track
382  * of the LVT offsets in software and reserve the offset for the same
383  * vector also to be used on other cores. An offset is freed by
384  * setting the entry to APIC_EILVT_MASKED.
385  *
386  * If the BIOS is right, there should be no conflicts. Otherwise a
387  * "[Firmware Bug]: ..." error message is generated. However, if
388  * software does not properly determines the offsets, it is not
389  * necessarily a BIOS bug.
390  */
391 
392 static atomic_t eilvt_offsets[APIC_EILVT_NR_MAX];
393 
394 static inline int eilvt_entry_is_changeable(unsigned int old, unsigned int new)
395 {
396 	return (old & APIC_EILVT_MASKED)
397 		|| (new == APIC_EILVT_MASKED)
398 		|| ((new & ~APIC_EILVT_MASKED) == old);
399 }
400 
401 static unsigned int reserve_eilvt_offset(int offset, unsigned int new)
402 {
403 	unsigned int rsvd, vector;
404 
405 	if (offset >= APIC_EILVT_NR_MAX)
406 		return ~0;
407 
408 	rsvd = atomic_read(&eilvt_offsets[offset]);
409 	do {
410 		vector = rsvd & ~APIC_EILVT_MASKED;	/* 0: unassigned */
411 		if (vector && !eilvt_entry_is_changeable(vector, new))
412 			/* may not change if vectors are different */
413 			return rsvd;
414 		rsvd = atomic_cmpxchg(&eilvt_offsets[offset], rsvd, new);
415 	} while (rsvd != new);
416 
417 	rsvd &= ~APIC_EILVT_MASKED;
418 	if (rsvd && rsvd != vector)
419 		pr_info("LVT offset %d assigned for vector 0x%02x\n",
420 			offset, rsvd);
421 
422 	return new;
423 }
424 
425 /*
426  * If mask=1, the LVT entry does not generate interrupts while mask=0
427  * enables the vector. See also the BKDGs. Must be called with
428  * preemption disabled.
429  */
430 
431 int setup_APIC_eilvt(u8 offset, u8 vector, u8 msg_type, u8 mask)
432 {
433 	unsigned long reg = APIC_EILVTn(offset);
434 	unsigned int new, old, reserved;
435 
436 	new = (mask << 16) | (msg_type << 8) | vector;
437 	old = apic_read(reg);
438 	reserved = reserve_eilvt_offset(offset, new);
439 
440 	if (reserved != new) {
441 		pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
442 		       "vector 0x%x, but the register is already in use for "
443 		       "vector 0x%x on another cpu\n",
444 		       smp_processor_id(), reg, offset, new, reserved);
445 		return -EINVAL;
446 	}
447 
448 	if (!eilvt_entry_is_changeable(old, new)) {
449 		pr_err(FW_BUG "cpu %d, try to use APIC%lX (LVT offset %d) for "
450 		       "vector 0x%x, but the register is already in use for "
451 		       "vector 0x%x on this cpu\n",
452 		       smp_processor_id(), reg, offset, new, old);
453 		return -EBUSY;
454 	}
455 
456 	apic_write(reg, new);
457 
458 	return 0;
459 }
460 EXPORT_SYMBOL_GPL(setup_APIC_eilvt);
461 
462 /*
463  * Program the next event, relative to now
464  */
465 static int lapic_next_event(unsigned long delta,
466 			    struct clock_event_device *evt)
467 {
468 	apic_write(APIC_TMICT, delta);
469 	return 0;
470 }
471 
472 static int lapic_next_deadline(unsigned long delta,
473 			       struct clock_event_device *evt)
474 {
475 	u64 tsc;
476 
477 	tsc = rdtsc();
478 	wrmsrl(MSR_IA32_TSC_DEADLINE, tsc + (((u64) delta) * TSC_DIVISOR));
479 	return 0;
480 }
481 
482 static int lapic_timer_shutdown(struct clock_event_device *evt)
483 {
484 	unsigned int v;
485 
486 	/* Lapic used as dummy for broadcast ? */
487 	if (evt->features & CLOCK_EVT_FEAT_DUMMY)
488 		return 0;
489 
490 	v = apic_read(APIC_LVTT);
491 	v |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
492 	apic_write(APIC_LVTT, v);
493 	apic_write(APIC_TMICT, 0);
494 	return 0;
495 }
496 
497 static inline int
498 lapic_timer_set_periodic_oneshot(struct clock_event_device *evt, bool oneshot)
499 {
500 	/* Lapic used as dummy for broadcast ? */
501 	if (evt->features & CLOCK_EVT_FEAT_DUMMY)
502 		return 0;
503 
504 	__setup_APIC_LVTT(lapic_timer_period, oneshot, 1);
505 	return 0;
506 }
507 
508 static int lapic_timer_set_periodic(struct clock_event_device *evt)
509 {
510 	return lapic_timer_set_periodic_oneshot(evt, false);
511 }
512 
513 static int lapic_timer_set_oneshot(struct clock_event_device *evt)
514 {
515 	return lapic_timer_set_periodic_oneshot(evt, true);
516 }
517 
518 /*
519  * Local APIC timer broadcast function
520  */
521 static void lapic_timer_broadcast(const struct cpumask *mask)
522 {
523 #ifdef CONFIG_SMP
524 	apic->send_IPI_mask(mask, LOCAL_TIMER_VECTOR);
525 #endif
526 }
527 
528 
529 /*
530  * The local apic timer can be used for any function which is CPU local.
531  */
532 static struct clock_event_device lapic_clockevent = {
533 	.name				= "lapic",
534 	.features			= CLOCK_EVT_FEAT_PERIODIC |
535 					  CLOCK_EVT_FEAT_ONESHOT | CLOCK_EVT_FEAT_C3STOP
536 					  | CLOCK_EVT_FEAT_DUMMY,
537 	.shift				= 32,
538 	.set_state_shutdown		= lapic_timer_shutdown,
539 	.set_state_periodic		= lapic_timer_set_periodic,
540 	.set_state_oneshot		= lapic_timer_set_oneshot,
541 	.set_state_oneshot_stopped	= lapic_timer_shutdown,
542 	.set_next_event			= lapic_next_event,
543 	.broadcast			= lapic_timer_broadcast,
544 	.rating				= 100,
545 	.irq				= -1,
546 };
547 static DEFINE_PER_CPU(struct clock_event_device, lapic_events);
548 
549 #define DEADLINE_MODEL_MATCH_FUNC(model, func)	\
550 	{ X86_VENDOR_INTEL, 6, model, X86_FEATURE_ANY, (unsigned long)&func }
551 
552 #define DEADLINE_MODEL_MATCH_REV(model, rev)	\
553 	{ X86_VENDOR_INTEL, 6, model, X86_FEATURE_ANY, (unsigned long)rev }
554 
555 static u32 hsx_deadline_rev(void)
556 {
557 	switch (boot_cpu_data.x86_stepping) {
558 	case 0x02: return 0x3a; /* EP */
559 	case 0x04: return 0x0f; /* EX */
560 	}
561 
562 	return ~0U;
563 }
564 
565 static u32 bdx_deadline_rev(void)
566 {
567 	switch (boot_cpu_data.x86_stepping) {
568 	case 0x02: return 0x00000011;
569 	case 0x03: return 0x0700000e;
570 	case 0x04: return 0x0f00000c;
571 	case 0x05: return 0x0e000003;
572 	}
573 
574 	return ~0U;
575 }
576 
577 static u32 skx_deadline_rev(void)
578 {
579 	switch (boot_cpu_data.x86_stepping) {
580 	case 0x03: return 0x01000136;
581 	case 0x04: return 0x02000014;
582 	}
583 
584 	if (boot_cpu_data.x86_stepping > 4)
585 		return 0;
586 
587 	return ~0U;
588 }
589 
590 static const struct x86_cpu_id deadline_match[] = {
591 	DEADLINE_MODEL_MATCH_FUNC( INTEL_FAM6_HASWELL_X,	hsx_deadline_rev),
592 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_BROADWELL_X,	0x0b000020),
593 	DEADLINE_MODEL_MATCH_FUNC( INTEL_FAM6_BROADWELL_D,	bdx_deadline_rev),
594 	DEADLINE_MODEL_MATCH_FUNC( INTEL_FAM6_SKYLAKE_X,	skx_deadline_rev),
595 
596 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_HASWELL,		0x22),
597 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_HASWELL_L,	0x20),
598 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_HASWELL_G,	0x17),
599 
600 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_BROADWELL,	0x25),
601 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_BROADWELL_G,	0x17),
602 
603 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_SKYLAKE_L,	0xb2),
604 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_SKYLAKE,		0xb2),
605 
606 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_KABYLAKE_L,	0x52),
607 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_KABYLAKE,		0x52),
608 
609 	{},
610 };
611 
612 static void apic_check_deadline_errata(void)
613 {
614 	const struct x86_cpu_id *m;
615 	u32 rev;
616 
617 	if (!boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER) ||
618 	    boot_cpu_has(X86_FEATURE_HYPERVISOR))
619 		return;
620 
621 	m = x86_match_cpu(deadline_match);
622 	if (!m)
623 		return;
624 
625 	/*
626 	 * Function pointers will have the MSB set due to address layout,
627 	 * immediate revisions will not.
628 	 */
629 	if ((long)m->driver_data < 0)
630 		rev = ((u32 (*)(void))(m->driver_data))();
631 	else
632 		rev = (u32)m->driver_data;
633 
634 	if (boot_cpu_data.microcode >= rev)
635 		return;
636 
637 	setup_clear_cpu_cap(X86_FEATURE_TSC_DEADLINE_TIMER);
638 	pr_err(FW_BUG "TSC_DEADLINE disabled due to Errata; "
639 	       "please update microcode to version: 0x%x (or later)\n", rev);
640 }
641 
642 /*
643  * Setup the local APIC timer for this CPU. Copy the initialized values
644  * of the boot CPU and register the clock event in the framework.
645  */
646 static void setup_APIC_timer(void)
647 {
648 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
649 
650 	if (this_cpu_has(X86_FEATURE_ARAT)) {
651 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_C3STOP;
652 		/* Make LAPIC timer preferrable over percpu HPET */
653 		lapic_clockevent.rating = 150;
654 	}
655 
656 	memcpy(levt, &lapic_clockevent, sizeof(*levt));
657 	levt->cpumask = cpumask_of(smp_processor_id());
658 
659 	if (this_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER)) {
660 		levt->name = "lapic-deadline";
661 		levt->features &= ~(CLOCK_EVT_FEAT_PERIODIC |
662 				    CLOCK_EVT_FEAT_DUMMY);
663 		levt->set_next_event = lapic_next_deadline;
664 		clockevents_config_and_register(levt,
665 						tsc_khz * (1000 / TSC_DIVISOR),
666 						0xF, ~0UL);
667 	} else
668 		clockevents_register_device(levt);
669 }
670 
671 /*
672  * Install the updated TSC frequency from recalibration at the TSC
673  * deadline clockevent devices.
674  */
675 static void __lapic_update_tsc_freq(void *info)
676 {
677 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
678 
679 	if (!this_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
680 		return;
681 
682 	clockevents_update_freq(levt, tsc_khz * (1000 / TSC_DIVISOR));
683 }
684 
685 void lapic_update_tsc_freq(void)
686 {
687 	/*
688 	 * The clockevent device's ->mult and ->shift can both be
689 	 * changed. In order to avoid races, schedule the frequency
690 	 * update code on each CPU.
691 	 */
692 	on_each_cpu(__lapic_update_tsc_freq, NULL, 0);
693 }
694 
695 /*
696  * In this functions we calibrate APIC bus clocks to the external timer.
697  *
698  * We want to do the calibration only once since we want to have local timer
699  * irqs syncron. CPUs connected by the same APIC bus have the very same bus
700  * frequency.
701  *
702  * This was previously done by reading the PIT/HPET and waiting for a wrap
703  * around to find out, that a tick has elapsed. I have a box, where the PIT
704  * readout is broken, so it never gets out of the wait loop again. This was
705  * also reported by others.
706  *
707  * Monitoring the jiffies value is inaccurate and the clockevents
708  * infrastructure allows us to do a simple substitution of the interrupt
709  * handler.
710  *
711  * The calibration routine also uses the pm_timer when possible, as the PIT
712  * happens to run way too slow (factor 2.3 on my VAIO CoreDuo, which goes
713  * back to normal later in the boot process).
714  */
715 
716 #define LAPIC_CAL_LOOPS		(HZ/10)
717 
718 static __initdata int lapic_cal_loops = -1;
719 static __initdata long lapic_cal_t1, lapic_cal_t2;
720 static __initdata unsigned long long lapic_cal_tsc1, lapic_cal_tsc2;
721 static __initdata unsigned long lapic_cal_pm1, lapic_cal_pm2;
722 static __initdata unsigned long lapic_cal_j1, lapic_cal_j2;
723 
724 /*
725  * Temporary interrupt handler and polled calibration function.
726  */
727 static void __init lapic_cal_handler(struct clock_event_device *dev)
728 {
729 	unsigned long long tsc = 0;
730 	long tapic = apic_read(APIC_TMCCT);
731 	unsigned long pm = acpi_pm_read_early();
732 
733 	if (boot_cpu_has(X86_FEATURE_TSC))
734 		tsc = rdtsc();
735 
736 	switch (lapic_cal_loops++) {
737 	case 0:
738 		lapic_cal_t1 = tapic;
739 		lapic_cal_tsc1 = tsc;
740 		lapic_cal_pm1 = pm;
741 		lapic_cal_j1 = jiffies;
742 		break;
743 
744 	case LAPIC_CAL_LOOPS:
745 		lapic_cal_t2 = tapic;
746 		lapic_cal_tsc2 = tsc;
747 		if (pm < lapic_cal_pm1)
748 			pm += ACPI_PM_OVRRUN;
749 		lapic_cal_pm2 = pm;
750 		lapic_cal_j2 = jiffies;
751 		break;
752 	}
753 }
754 
755 static int __init
756 calibrate_by_pmtimer(long deltapm, long *delta, long *deltatsc)
757 {
758 	const long pm_100ms = PMTMR_TICKS_PER_SEC / 10;
759 	const long pm_thresh = pm_100ms / 100;
760 	unsigned long mult;
761 	u64 res;
762 
763 #ifndef CONFIG_X86_PM_TIMER
764 	return -1;
765 #endif
766 
767 	apic_printk(APIC_VERBOSE, "... PM-Timer delta = %ld\n", deltapm);
768 
769 	/* Check, if the PM timer is available */
770 	if (!deltapm)
771 		return -1;
772 
773 	mult = clocksource_hz2mult(PMTMR_TICKS_PER_SEC, 22);
774 
775 	if (deltapm > (pm_100ms - pm_thresh) &&
776 	    deltapm < (pm_100ms + pm_thresh)) {
777 		apic_printk(APIC_VERBOSE, "... PM-Timer result ok\n");
778 		return 0;
779 	}
780 
781 	res = (((u64)deltapm) *  mult) >> 22;
782 	do_div(res, 1000000);
783 	pr_warn("APIC calibration not consistent "
784 		"with PM-Timer: %ldms instead of 100ms\n", (long)res);
785 
786 	/* Correct the lapic counter value */
787 	res = (((u64)(*delta)) * pm_100ms);
788 	do_div(res, deltapm);
789 	pr_info("APIC delta adjusted to PM-Timer: "
790 		"%lu (%ld)\n", (unsigned long)res, *delta);
791 	*delta = (long)res;
792 
793 	/* Correct the tsc counter value */
794 	if (boot_cpu_has(X86_FEATURE_TSC)) {
795 		res = (((u64)(*deltatsc)) * pm_100ms);
796 		do_div(res, deltapm);
797 		apic_printk(APIC_VERBOSE, "TSC delta adjusted to "
798 					  "PM-Timer: %lu (%ld)\n",
799 					(unsigned long)res, *deltatsc);
800 		*deltatsc = (long)res;
801 	}
802 
803 	return 0;
804 }
805 
806 static int __init lapic_init_clockevent(void)
807 {
808 	if (!lapic_timer_period)
809 		return -1;
810 
811 	/* Calculate the scaled math multiplication factor */
812 	lapic_clockevent.mult = div_sc(lapic_timer_period/APIC_DIVISOR,
813 					TICK_NSEC, lapic_clockevent.shift);
814 	lapic_clockevent.max_delta_ns =
815 		clockevent_delta2ns(0x7FFFFFFF, &lapic_clockevent);
816 	lapic_clockevent.max_delta_ticks = 0x7FFFFFFF;
817 	lapic_clockevent.min_delta_ns =
818 		clockevent_delta2ns(0xF, &lapic_clockevent);
819 	lapic_clockevent.min_delta_ticks = 0xF;
820 
821 	return 0;
822 }
823 
824 bool __init apic_needs_pit(void)
825 {
826 	/*
827 	 * If the frequencies are not known, PIT is required for both TSC
828 	 * and apic timer calibration.
829 	 */
830 	if (!tsc_khz || !cpu_khz)
831 		return true;
832 
833 	/* Is there an APIC at all or is it disabled? */
834 	if (!boot_cpu_has(X86_FEATURE_APIC) || disable_apic)
835 		return true;
836 
837 	/*
838 	 * If interrupt delivery mode is legacy PIC or virtual wire without
839 	 * configuration, the local APIC timer wont be set up. Make sure
840 	 * that the PIT is initialized.
841 	 */
842 	if (apic_intr_mode == APIC_PIC ||
843 	    apic_intr_mode == APIC_VIRTUAL_WIRE_NO_CONFIG)
844 		return true;
845 
846 	/* Virt guests may lack ARAT, but still have DEADLINE */
847 	if (!boot_cpu_has(X86_FEATURE_ARAT))
848 		return true;
849 
850 	/* Deadline timer is based on TSC so no further PIT action required */
851 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
852 		return false;
853 
854 	/* APIC timer disabled? */
855 	if (disable_apic_timer)
856 		return true;
857 	/*
858 	 * The APIC timer frequency is known already, no PIT calibration
859 	 * required. If unknown, let the PIT be initialized.
860 	 */
861 	return lapic_timer_period == 0;
862 }
863 
864 static int __init calibrate_APIC_clock(void)
865 {
866 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
867 	u64 tsc_perj = 0, tsc_start = 0;
868 	unsigned long jif_start;
869 	unsigned long deltaj;
870 	long delta, deltatsc;
871 	int pm_referenced = 0;
872 
873 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
874 		return 0;
875 
876 	/*
877 	 * Check if lapic timer has already been calibrated by platform
878 	 * specific routine, such as tsc calibration code. If so just fill
879 	 * in the clockevent structure and return.
880 	 */
881 	if (!lapic_init_clockevent()) {
882 		apic_printk(APIC_VERBOSE, "lapic timer already calibrated %d\n",
883 			    lapic_timer_period);
884 		/*
885 		 * Direct calibration methods must have an always running
886 		 * local APIC timer, no need for broadcast timer.
887 		 */
888 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
889 		return 0;
890 	}
891 
892 	apic_printk(APIC_VERBOSE, "Using local APIC timer interrupts.\n"
893 		    "calibrating APIC timer ...\n");
894 
895 	/*
896 	 * There are platforms w/o global clockevent devices. Instead of
897 	 * making the calibration conditional on that, use a polling based
898 	 * approach everywhere.
899 	 */
900 	local_irq_disable();
901 
902 	/*
903 	 * Setup the APIC counter to maximum. There is no way the lapic
904 	 * can underflow in the 100ms detection time frame
905 	 */
906 	__setup_APIC_LVTT(0xffffffff, 0, 0);
907 
908 	/*
909 	 * Methods to terminate the calibration loop:
910 	 *  1) Global clockevent if available (jiffies)
911 	 *  2) TSC if available and frequency is known
912 	 */
913 	jif_start = READ_ONCE(jiffies);
914 
915 	if (tsc_khz) {
916 		tsc_start = rdtsc();
917 		tsc_perj = div_u64((u64)tsc_khz * 1000, HZ);
918 	}
919 
920 	/*
921 	 * Enable interrupts so the tick can fire, if a global
922 	 * clockevent device is available
923 	 */
924 	local_irq_enable();
925 
926 	while (lapic_cal_loops <= LAPIC_CAL_LOOPS) {
927 		/* Wait for a tick to elapse */
928 		while (1) {
929 			if (tsc_khz) {
930 				u64 tsc_now = rdtsc();
931 				if ((tsc_now - tsc_start) >= tsc_perj) {
932 					tsc_start += tsc_perj;
933 					break;
934 				}
935 			} else {
936 				unsigned long jif_now = READ_ONCE(jiffies);
937 
938 				if (time_after(jif_now, jif_start)) {
939 					jif_start = jif_now;
940 					break;
941 				}
942 			}
943 			cpu_relax();
944 		}
945 
946 		/* Invoke the calibration routine */
947 		local_irq_disable();
948 		lapic_cal_handler(NULL);
949 		local_irq_enable();
950 	}
951 
952 	local_irq_disable();
953 
954 	/* Build delta t1-t2 as apic timer counts down */
955 	delta = lapic_cal_t1 - lapic_cal_t2;
956 	apic_printk(APIC_VERBOSE, "... lapic delta = %ld\n", delta);
957 
958 	deltatsc = (long)(lapic_cal_tsc2 - lapic_cal_tsc1);
959 
960 	/* we trust the PM based calibration if possible */
961 	pm_referenced = !calibrate_by_pmtimer(lapic_cal_pm2 - lapic_cal_pm1,
962 					&delta, &deltatsc);
963 
964 	lapic_timer_period = (delta * APIC_DIVISOR) / LAPIC_CAL_LOOPS;
965 	lapic_init_clockevent();
966 
967 	apic_printk(APIC_VERBOSE, "..... delta %ld\n", delta);
968 	apic_printk(APIC_VERBOSE, "..... mult: %u\n", lapic_clockevent.mult);
969 	apic_printk(APIC_VERBOSE, "..... calibration result: %u\n",
970 		    lapic_timer_period);
971 
972 	if (boot_cpu_has(X86_FEATURE_TSC)) {
973 		apic_printk(APIC_VERBOSE, "..... CPU clock speed is "
974 			    "%ld.%04ld MHz.\n",
975 			    (deltatsc / LAPIC_CAL_LOOPS) / (1000000 / HZ),
976 			    (deltatsc / LAPIC_CAL_LOOPS) % (1000000 / HZ));
977 	}
978 
979 	apic_printk(APIC_VERBOSE, "..... host bus clock speed is "
980 		    "%u.%04u MHz.\n",
981 		    lapic_timer_period / (1000000 / HZ),
982 		    lapic_timer_period % (1000000 / HZ));
983 
984 	/*
985 	 * Do a sanity check on the APIC calibration result
986 	 */
987 	if (lapic_timer_period < (1000000 / HZ)) {
988 		local_irq_enable();
989 		pr_warn("APIC frequency too slow, disabling apic timer\n");
990 		return -1;
991 	}
992 
993 	levt->features &= ~CLOCK_EVT_FEAT_DUMMY;
994 
995 	/*
996 	 * PM timer calibration failed or not turned on so lets try APIC
997 	 * timer based calibration, if a global clockevent device is
998 	 * available.
999 	 */
1000 	if (!pm_referenced && global_clock_event) {
1001 		apic_printk(APIC_VERBOSE, "... verify APIC timer\n");
1002 
1003 		/*
1004 		 * Setup the apic timer manually
1005 		 */
1006 		levt->event_handler = lapic_cal_handler;
1007 		lapic_timer_set_periodic(levt);
1008 		lapic_cal_loops = -1;
1009 
1010 		/* Let the interrupts run */
1011 		local_irq_enable();
1012 
1013 		while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
1014 			cpu_relax();
1015 
1016 		/* Stop the lapic timer */
1017 		local_irq_disable();
1018 		lapic_timer_shutdown(levt);
1019 
1020 		/* Jiffies delta */
1021 		deltaj = lapic_cal_j2 - lapic_cal_j1;
1022 		apic_printk(APIC_VERBOSE, "... jiffies delta = %lu\n", deltaj);
1023 
1024 		/* Check, if the jiffies result is consistent */
1025 		if (deltaj >= LAPIC_CAL_LOOPS-2 && deltaj <= LAPIC_CAL_LOOPS+2)
1026 			apic_printk(APIC_VERBOSE, "... jiffies result ok\n");
1027 		else
1028 			levt->features |= CLOCK_EVT_FEAT_DUMMY;
1029 	}
1030 	local_irq_enable();
1031 
1032 	if (levt->features & CLOCK_EVT_FEAT_DUMMY) {
1033 		pr_warn("APIC timer disabled due to verification failure\n");
1034 		return -1;
1035 	}
1036 
1037 	return 0;
1038 }
1039 
1040 /*
1041  * Setup the boot APIC
1042  *
1043  * Calibrate and verify the result.
1044  */
1045 void __init setup_boot_APIC_clock(void)
1046 {
1047 	/*
1048 	 * The local apic timer can be disabled via the kernel
1049 	 * commandline or from the CPU detection code. Register the lapic
1050 	 * timer as a dummy clock event source on SMP systems, so the
1051 	 * broadcast mechanism is used. On UP systems simply ignore it.
1052 	 */
1053 	if (disable_apic_timer) {
1054 		pr_info("Disabling APIC timer\n");
1055 		/* No broadcast on UP ! */
1056 		if (num_possible_cpus() > 1) {
1057 			lapic_clockevent.mult = 1;
1058 			setup_APIC_timer();
1059 		}
1060 		return;
1061 	}
1062 
1063 	if (calibrate_APIC_clock()) {
1064 		/* No broadcast on UP ! */
1065 		if (num_possible_cpus() > 1)
1066 			setup_APIC_timer();
1067 		return;
1068 	}
1069 
1070 	/*
1071 	 * If nmi_watchdog is set to IO_APIC, we need the
1072 	 * PIT/HPET going.  Otherwise register lapic as a dummy
1073 	 * device.
1074 	 */
1075 	lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
1076 
1077 	/* Setup the lapic or request the broadcast */
1078 	setup_APIC_timer();
1079 	amd_e400_c1e_apic_setup();
1080 }
1081 
1082 void setup_secondary_APIC_clock(void)
1083 {
1084 	setup_APIC_timer();
1085 	amd_e400_c1e_apic_setup();
1086 }
1087 
1088 /*
1089  * The guts of the apic timer interrupt
1090  */
1091 static void local_apic_timer_interrupt(void)
1092 {
1093 	struct clock_event_device *evt = this_cpu_ptr(&lapic_events);
1094 
1095 	/*
1096 	 * Normally we should not be here till LAPIC has been initialized but
1097 	 * in some cases like kdump, its possible that there is a pending LAPIC
1098 	 * timer interrupt from previous kernel's context and is delivered in
1099 	 * new kernel the moment interrupts are enabled.
1100 	 *
1101 	 * Interrupts are enabled early and LAPIC is setup much later, hence
1102 	 * its possible that when we get here evt->event_handler is NULL.
1103 	 * Check for event_handler being NULL and discard the interrupt as
1104 	 * spurious.
1105 	 */
1106 	if (!evt->event_handler) {
1107 		pr_warn("Spurious LAPIC timer interrupt on cpu %d\n",
1108 			smp_processor_id());
1109 		/* Switch it off */
1110 		lapic_timer_shutdown(evt);
1111 		return;
1112 	}
1113 
1114 	/*
1115 	 * the NMI deadlock-detector uses this.
1116 	 */
1117 	inc_irq_stat(apic_timer_irqs);
1118 
1119 	evt->event_handler(evt);
1120 }
1121 
1122 /*
1123  * Local APIC timer interrupt. This is the most natural way for doing
1124  * local interrupts, but local timer interrupts can be emulated by
1125  * broadcast interrupts too. [in case the hw doesn't support APIC timers]
1126  *
1127  * [ if a single-CPU system runs an SMP kernel then we call the local
1128  *   interrupt as well. Thus we cannot inline the local irq ... ]
1129  */
1130 __visible void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
1131 {
1132 	struct pt_regs *old_regs = set_irq_regs(regs);
1133 
1134 	/*
1135 	 * NOTE! We'd better ACK the irq immediately,
1136 	 * because timer handling can be slow.
1137 	 *
1138 	 * update_process_times() expects us to have done irq_enter().
1139 	 * Besides, if we don't timer interrupts ignore the global
1140 	 * interrupt lock, which is the WrongThing (tm) to do.
1141 	 */
1142 	entering_ack_irq();
1143 	trace_local_timer_entry(LOCAL_TIMER_VECTOR);
1144 	local_apic_timer_interrupt();
1145 	trace_local_timer_exit(LOCAL_TIMER_VECTOR);
1146 	exiting_irq();
1147 
1148 	set_irq_regs(old_regs);
1149 }
1150 
1151 int setup_profiling_timer(unsigned int multiplier)
1152 {
1153 	return -EINVAL;
1154 }
1155 
1156 /*
1157  * Local APIC start and shutdown
1158  */
1159 
1160 /**
1161  * clear_local_APIC - shutdown the local APIC
1162  *
1163  * This is called, when a CPU is disabled and before rebooting, so the state of
1164  * the local APIC has no dangling leftovers. Also used to cleanout any BIOS
1165  * leftovers during boot.
1166  */
1167 void clear_local_APIC(void)
1168 {
1169 	int maxlvt;
1170 	u32 v;
1171 
1172 	/* APIC hasn't been mapped yet */
1173 	if (!x2apic_mode && !apic_phys)
1174 		return;
1175 
1176 	maxlvt = lapic_get_maxlvt();
1177 	/*
1178 	 * Masking an LVT entry can trigger a local APIC error
1179 	 * if the vector is zero. Mask LVTERR first to prevent this.
1180 	 */
1181 	if (maxlvt >= 3) {
1182 		v = ERROR_APIC_VECTOR; /* any non-zero vector will do */
1183 		apic_write(APIC_LVTERR, v | APIC_LVT_MASKED);
1184 	}
1185 	/*
1186 	 * Careful: we have to set masks only first to deassert
1187 	 * any level-triggered sources.
1188 	 */
1189 	v = apic_read(APIC_LVTT);
1190 	apic_write(APIC_LVTT, v | APIC_LVT_MASKED);
1191 	v = apic_read(APIC_LVT0);
1192 	apic_write(APIC_LVT0, v | APIC_LVT_MASKED);
1193 	v = apic_read(APIC_LVT1);
1194 	apic_write(APIC_LVT1, v | APIC_LVT_MASKED);
1195 	if (maxlvt >= 4) {
1196 		v = apic_read(APIC_LVTPC);
1197 		apic_write(APIC_LVTPC, v | APIC_LVT_MASKED);
1198 	}
1199 
1200 	/* lets not touch this if we didn't frob it */
1201 #ifdef CONFIG_X86_THERMAL_VECTOR
1202 	if (maxlvt >= 5) {
1203 		v = apic_read(APIC_LVTTHMR);
1204 		apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED);
1205 	}
1206 #endif
1207 #ifdef CONFIG_X86_MCE_INTEL
1208 	if (maxlvt >= 6) {
1209 		v = apic_read(APIC_LVTCMCI);
1210 		if (!(v & APIC_LVT_MASKED))
1211 			apic_write(APIC_LVTCMCI, v | APIC_LVT_MASKED);
1212 	}
1213 #endif
1214 
1215 	/*
1216 	 * Clean APIC state for other OSs:
1217 	 */
1218 	apic_write(APIC_LVTT, APIC_LVT_MASKED);
1219 	apic_write(APIC_LVT0, APIC_LVT_MASKED);
1220 	apic_write(APIC_LVT1, APIC_LVT_MASKED);
1221 	if (maxlvt >= 3)
1222 		apic_write(APIC_LVTERR, APIC_LVT_MASKED);
1223 	if (maxlvt >= 4)
1224 		apic_write(APIC_LVTPC, APIC_LVT_MASKED);
1225 
1226 	/* Integrated APIC (!82489DX) ? */
1227 	if (lapic_is_integrated()) {
1228 		if (maxlvt > 3)
1229 			/* Clear ESR due to Pentium errata 3AP and 11AP */
1230 			apic_write(APIC_ESR, 0);
1231 		apic_read(APIC_ESR);
1232 	}
1233 }
1234 
1235 /**
1236  * apic_soft_disable - Clears and software disables the local APIC on hotplug
1237  *
1238  * Contrary to disable_local_APIC() this does not touch the enable bit in
1239  * MSR_IA32_APICBASE. Clearing that bit on systems based on the 3 wire APIC
1240  * bus would require a hardware reset as the APIC would lose track of bus
1241  * arbitration. On systems with FSB delivery APICBASE could be disabled,
1242  * but it has to be guaranteed that no interrupt is sent to the APIC while
1243  * in that state and it's not clear from the SDM whether it still responds
1244  * to INIT/SIPI messages. Stay on the safe side and use software disable.
1245  */
1246 void apic_soft_disable(void)
1247 {
1248 	u32 value;
1249 
1250 	clear_local_APIC();
1251 
1252 	/* Soft disable APIC (implies clearing of registers for 82489DX!). */
1253 	value = apic_read(APIC_SPIV);
1254 	value &= ~APIC_SPIV_APIC_ENABLED;
1255 	apic_write(APIC_SPIV, value);
1256 }
1257 
1258 /**
1259  * disable_local_APIC - clear and disable the local APIC
1260  */
1261 void disable_local_APIC(void)
1262 {
1263 	/* APIC hasn't been mapped yet */
1264 	if (!x2apic_mode && !apic_phys)
1265 		return;
1266 
1267 	apic_soft_disable();
1268 
1269 #ifdef CONFIG_X86_32
1270 	/*
1271 	 * When LAPIC was disabled by the BIOS and enabled by the kernel,
1272 	 * restore the disabled state.
1273 	 */
1274 	if (enabled_via_apicbase) {
1275 		unsigned int l, h;
1276 
1277 		rdmsr(MSR_IA32_APICBASE, l, h);
1278 		l &= ~MSR_IA32_APICBASE_ENABLE;
1279 		wrmsr(MSR_IA32_APICBASE, l, h);
1280 	}
1281 #endif
1282 }
1283 
1284 /*
1285  * If Linux enabled the LAPIC against the BIOS default disable it down before
1286  * re-entering the BIOS on shutdown.  Otherwise the BIOS may get confused and
1287  * not power-off.  Additionally clear all LVT entries before disable_local_APIC
1288  * for the case where Linux didn't enable the LAPIC.
1289  */
1290 void lapic_shutdown(void)
1291 {
1292 	unsigned long flags;
1293 
1294 	if (!boot_cpu_has(X86_FEATURE_APIC) && !apic_from_smp_config())
1295 		return;
1296 
1297 	local_irq_save(flags);
1298 
1299 #ifdef CONFIG_X86_32
1300 	if (!enabled_via_apicbase)
1301 		clear_local_APIC();
1302 	else
1303 #endif
1304 		disable_local_APIC();
1305 
1306 
1307 	local_irq_restore(flags);
1308 }
1309 
1310 /**
1311  * sync_Arb_IDs - synchronize APIC bus arbitration IDs
1312  */
1313 void __init sync_Arb_IDs(void)
1314 {
1315 	/*
1316 	 * Unsupported on P4 - see Intel Dev. Manual Vol. 3, Ch. 8.6.1 And not
1317 	 * needed on AMD.
1318 	 */
1319 	if (modern_apic() || boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
1320 		return;
1321 
1322 	/*
1323 	 * Wait for idle.
1324 	 */
1325 	apic_wait_icr_idle();
1326 
1327 	apic_printk(APIC_DEBUG, "Synchronizing Arb IDs.\n");
1328 	apic_write(APIC_ICR, APIC_DEST_ALLINC |
1329 			APIC_INT_LEVELTRIG | APIC_DM_INIT);
1330 }
1331 
1332 enum apic_intr_mode_id apic_intr_mode __ro_after_init;
1333 
1334 static int __init __apic_intr_mode_select(void)
1335 {
1336 	/* Check kernel option */
1337 	if (disable_apic) {
1338 		pr_info("APIC disabled via kernel command line\n");
1339 		return APIC_PIC;
1340 	}
1341 
1342 	/* Check BIOS */
1343 #ifdef CONFIG_X86_64
1344 	/* On 64-bit, the APIC must be integrated, Check local APIC only */
1345 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1346 		disable_apic = 1;
1347 		pr_info("APIC disabled by BIOS\n");
1348 		return APIC_PIC;
1349 	}
1350 #else
1351 	/* On 32-bit, the APIC may be integrated APIC or 82489DX */
1352 
1353 	/* Neither 82489DX nor integrated APIC ? */
1354 	if (!boot_cpu_has(X86_FEATURE_APIC) && !smp_found_config) {
1355 		disable_apic = 1;
1356 		return APIC_PIC;
1357 	}
1358 
1359 	/* If the BIOS pretends there is an integrated APIC ? */
1360 	if (!boot_cpu_has(X86_FEATURE_APIC) &&
1361 		APIC_INTEGRATED(boot_cpu_apic_version)) {
1362 		disable_apic = 1;
1363 		pr_err(FW_BUG "Local APIC %d not detected, force emulation\n",
1364 				       boot_cpu_physical_apicid);
1365 		return APIC_PIC;
1366 	}
1367 #endif
1368 
1369 	/* Check MP table or ACPI MADT configuration */
1370 	if (!smp_found_config) {
1371 		disable_ioapic_support();
1372 		if (!acpi_lapic) {
1373 			pr_info("APIC: ACPI MADT or MP tables are not detected\n");
1374 			return APIC_VIRTUAL_WIRE_NO_CONFIG;
1375 		}
1376 		return APIC_VIRTUAL_WIRE;
1377 	}
1378 
1379 #ifdef CONFIG_SMP
1380 	/* If SMP should be disabled, then really disable it! */
1381 	if (!setup_max_cpus) {
1382 		pr_info("APIC: SMP mode deactivated\n");
1383 		return APIC_SYMMETRIC_IO_NO_ROUTING;
1384 	}
1385 
1386 	if (read_apic_id() != boot_cpu_physical_apicid) {
1387 		panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
1388 		     read_apic_id(), boot_cpu_physical_apicid);
1389 		/* Or can we switch back to PIC here? */
1390 	}
1391 #endif
1392 
1393 	return APIC_SYMMETRIC_IO;
1394 }
1395 
1396 /* Select the interrupt delivery mode for the BSP */
1397 void __init apic_intr_mode_select(void)
1398 {
1399 	apic_intr_mode = __apic_intr_mode_select();
1400 }
1401 
1402 /*
1403  * An initial setup of the virtual wire mode.
1404  */
1405 void __init init_bsp_APIC(void)
1406 {
1407 	unsigned int value;
1408 
1409 	/*
1410 	 * Don't do the setup now if we have a SMP BIOS as the
1411 	 * through-I/O-APIC virtual wire mode might be active.
1412 	 */
1413 	if (smp_found_config || !boot_cpu_has(X86_FEATURE_APIC))
1414 		return;
1415 
1416 	/*
1417 	 * Do not trust the local APIC being empty at bootup.
1418 	 */
1419 	clear_local_APIC();
1420 
1421 	/*
1422 	 * Enable APIC.
1423 	 */
1424 	value = apic_read(APIC_SPIV);
1425 	value &= ~APIC_VECTOR_MASK;
1426 	value |= APIC_SPIV_APIC_ENABLED;
1427 
1428 #ifdef CONFIG_X86_32
1429 	/* This bit is reserved on P4/Xeon and should be cleared */
1430 	if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
1431 	    (boot_cpu_data.x86 == 15))
1432 		value &= ~APIC_SPIV_FOCUS_DISABLED;
1433 	else
1434 #endif
1435 		value |= APIC_SPIV_FOCUS_DISABLED;
1436 	value |= SPURIOUS_APIC_VECTOR;
1437 	apic_write(APIC_SPIV, value);
1438 
1439 	/*
1440 	 * Set up the virtual wire mode.
1441 	 */
1442 	apic_write(APIC_LVT0, APIC_DM_EXTINT);
1443 	value = APIC_DM_NMI;
1444 	if (!lapic_is_integrated())		/* 82489DX */
1445 		value |= APIC_LVT_LEVEL_TRIGGER;
1446 	if (apic_extnmi == APIC_EXTNMI_NONE)
1447 		value |= APIC_LVT_MASKED;
1448 	apic_write(APIC_LVT1, value);
1449 }
1450 
1451 static void __init apic_bsp_setup(bool upmode);
1452 
1453 /* Init the interrupt delivery mode for the BSP */
1454 void __init apic_intr_mode_init(void)
1455 {
1456 	bool upmode = IS_ENABLED(CONFIG_UP_LATE_INIT);
1457 
1458 	switch (apic_intr_mode) {
1459 	case APIC_PIC:
1460 		pr_info("APIC: Keep in PIC mode(8259)\n");
1461 		return;
1462 	case APIC_VIRTUAL_WIRE:
1463 		pr_info("APIC: Switch to virtual wire mode setup\n");
1464 		default_setup_apic_routing();
1465 		break;
1466 	case APIC_VIRTUAL_WIRE_NO_CONFIG:
1467 		pr_info("APIC: Switch to virtual wire mode setup with no configuration\n");
1468 		upmode = true;
1469 		default_setup_apic_routing();
1470 		break;
1471 	case APIC_SYMMETRIC_IO:
1472 		pr_info("APIC: Switch to symmetric I/O mode setup\n");
1473 		default_setup_apic_routing();
1474 		break;
1475 	case APIC_SYMMETRIC_IO_NO_ROUTING:
1476 		pr_info("APIC: Switch to symmetric I/O mode setup in no SMP routine\n");
1477 		break;
1478 	}
1479 
1480 	apic_bsp_setup(upmode);
1481 }
1482 
1483 static void lapic_setup_esr(void)
1484 {
1485 	unsigned int oldvalue, value, maxlvt;
1486 
1487 	if (!lapic_is_integrated()) {
1488 		pr_info("No ESR for 82489DX.\n");
1489 		return;
1490 	}
1491 
1492 	if (apic->disable_esr) {
1493 		/*
1494 		 * Something untraceable is creating bad interrupts on
1495 		 * secondary quads ... for the moment, just leave the
1496 		 * ESR disabled - we can't do anything useful with the
1497 		 * errors anyway - mbligh
1498 		 */
1499 		pr_info("Leaving ESR disabled.\n");
1500 		return;
1501 	}
1502 
1503 	maxlvt = lapic_get_maxlvt();
1504 	if (maxlvt > 3)		/* Due to the Pentium erratum 3AP. */
1505 		apic_write(APIC_ESR, 0);
1506 	oldvalue = apic_read(APIC_ESR);
1507 
1508 	/* enables sending errors */
1509 	value = ERROR_APIC_VECTOR;
1510 	apic_write(APIC_LVTERR, value);
1511 
1512 	/*
1513 	 * spec says clear errors after enabling vector.
1514 	 */
1515 	if (maxlvt > 3)
1516 		apic_write(APIC_ESR, 0);
1517 	value = apic_read(APIC_ESR);
1518 	if (value != oldvalue)
1519 		apic_printk(APIC_VERBOSE, "ESR value before enabling "
1520 			"vector: 0x%08x  after: 0x%08x\n",
1521 			oldvalue, value);
1522 }
1523 
1524 #define APIC_IR_REGS		APIC_ISR_NR
1525 #define APIC_IR_BITS		(APIC_IR_REGS * 32)
1526 #define APIC_IR_MAPSIZE		(APIC_IR_BITS / BITS_PER_LONG)
1527 
1528 union apic_ir {
1529 	unsigned long	map[APIC_IR_MAPSIZE];
1530 	u32		regs[APIC_IR_REGS];
1531 };
1532 
1533 static bool apic_check_and_ack(union apic_ir *irr, union apic_ir *isr)
1534 {
1535 	int i, bit;
1536 
1537 	/* Read the IRRs */
1538 	for (i = 0; i < APIC_IR_REGS; i++)
1539 		irr->regs[i] = apic_read(APIC_IRR + i * 0x10);
1540 
1541 	/* Read the ISRs */
1542 	for (i = 0; i < APIC_IR_REGS; i++)
1543 		isr->regs[i] = apic_read(APIC_ISR + i * 0x10);
1544 
1545 	/*
1546 	 * If the ISR map is not empty. ACK the APIC and run another round
1547 	 * to verify whether a pending IRR has been unblocked and turned
1548 	 * into a ISR.
1549 	 */
1550 	if (!bitmap_empty(isr->map, APIC_IR_BITS)) {
1551 		/*
1552 		 * There can be multiple ISR bits set when a high priority
1553 		 * interrupt preempted a lower priority one. Issue an ACK
1554 		 * per set bit.
1555 		 */
1556 		for_each_set_bit(bit, isr->map, APIC_IR_BITS)
1557 			ack_APIC_irq();
1558 		return true;
1559 	}
1560 
1561 	return !bitmap_empty(irr->map, APIC_IR_BITS);
1562 }
1563 
1564 /*
1565  * After a crash, we no longer service the interrupts and a pending
1566  * interrupt from previous kernel might still have ISR bit set.
1567  *
1568  * Most probably by now the CPU has serviced that pending interrupt and it
1569  * might not have done the ack_APIC_irq() because it thought, interrupt
1570  * came from i8259 as ExtInt. LAPIC did not get EOI so it does not clear
1571  * the ISR bit and cpu thinks it has already serivced the interrupt. Hence
1572  * a vector might get locked. It was noticed for timer irq (vector
1573  * 0x31). Issue an extra EOI to clear ISR.
1574  *
1575  * If there are pending IRR bits they turn into ISR bits after a higher
1576  * priority ISR bit has been acked.
1577  */
1578 static void apic_pending_intr_clear(void)
1579 {
1580 	union apic_ir irr, isr;
1581 	unsigned int i;
1582 
1583 	/* 512 loops are way oversized and give the APIC a chance to obey. */
1584 	for (i = 0; i < 512; i++) {
1585 		if (!apic_check_and_ack(&irr, &isr))
1586 			return;
1587 	}
1588 	/* Dump the IRR/ISR content if that failed */
1589 	pr_warn("APIC: Stale IRR: %256pb ISR: %256pb\n", irr.map, isr.map);
1590 }
1591 
1592 /**
1593  * setup_local_APIC - setup the local APIC
1594  *
1595  * Used to setup local APIC while initializing BSP or bringing up APs.
1596  * Always called with preemption disabled.
1597  */
1598 static void setup_local_APIC(void)
1599 {
1600 	int cpu = smp_processor_id();
1601 	unsigned int value;
1602 
1603 	if (disable_apic) {
1604 		disable_ioapic_support();
1605 		return;
1606 	}
1607 
1608 	/*
1609 	 * If this comes from kexec/kcrash the APIC might be enabled in
1610 	 * SPIV. Soft disable it before doing further initialization.
1611 	 */
1612 	value = apic_read(APIC_SPIV);
1613 	value &= ~APIC_SPIV_APIC_ENABLED;
1614 	apic_write(APIC_SPIV, value);
1615 
1616 #ifdef CONFIG_X86_32
1617 	/* Pound the ESR really hard over the head with a big hammer - mbligh */
1618 	if (lapic_is_integrated() && apic->disable_esr) {
1619 		apic_write(APIC_ESR, 0);
1620 		apic_write(APIC_ESR, 0);
1621 		apic_write(APIC_ESR, 0);
1622 		apic_write(APIC_ESR, 0);
1623 	}
1624 #endif
1625 	/*
1626 	 * Double-check whether this APIC is really registered.
1627 	 * This is meaningless in clustered apic mode, so we skip it.
1628 	 */
1629 	BUG_ON(!apic->apic_id_registered());
1630 
1631 	/*
1632 	 * Intel recommends to set DFR, LDR and TPR before enabling
1633 	 * an APIC.  See e.g. "AP-388 82489DX User's Manual" (Intel
1634 	 * document number 292116).  So here it goes...
1635 	 */
1636 	apic->init_apic_ldr();
1637 
1638 #ifdef CONFIG_X86_32
1639 	if (apic->dest_logical) {
1640 		int logical_apicid, ldr_apicid;
1641 
1642 		/*
1643 		 * APIC LDR is initialized.  If logical_apicid mapping was
1644 		 * initialized during get_smp_config(), make sure it matches
1645 		 * the actual value.
1646 		 */
1647 		logical_apicid = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
1648 		ldr_apicid = GET_APIC_LOGICAL_ID(apic_read(APIC_LDR));
1649 		if (logical_apicid != BAD_APICID)
1650 			WARN_ON(logical_apicid != ldr_apicid);
1651 		/* Always use the value from LDR. */
1652 		early_per_cpu(x86_cpu_to_logical_apicid, cpu) = ldr_apicid;
1653 	}
1654 #endif
1655 
1656 	/*
1657 	 * Set Task Priority to 'accept all except vectors 0-31'.  An APIC
1658 	 * vector in the 16-31 range could be delivered if TPR == 0, but we
1659 	 * would think it's an exception and terrible things will happen.  We
1660 	 * never change this later on.
1661 	 */
1662 	value = apic_read(APIC_TASKPRI);
1663 	value &= ~APIC_TPRI_MASK;
1664 	value |= 0x10;
1665 	apic_write(APIC_TASKPRI, value);
1666 
1667 	/* Clear eventually stale ISR/IRR bits */
1668 	apic_pending_intr_clear();
1669 
1670 	/*
1671 	 * Now that we are all set up, enable the APIC
1672 	 */
1673 	value = apic_read(APIC_SPIV);
1674 	value &= ~APIC_VECTOR_MASK;
1675 	/*
1676 	 * Enable APIC
1677 	 */
1678 	value |= APIC_SPIV_APIC_ENABLED;
1679 
1680 #ifdef CONFIG_X86_32
1681 	/*
1682 	 * Some unknown Intel IO/APIC (or APIC) errata is biting us with
1683 	 * certain networking cards. If high frequency interrupts are
1684 	 * happening on a particular IOAPIC pin, plus the IOAPIC routing
1685 	 * entry is masked/unmasked at a high rate as well then sooner or
1686 	 * later IOAPIC line gets 'stuck', no more interrupts are received
1687 	 * from the device. If focus CPU is disabled then the hang goes
1688 	 * away, oh well :-(
1689 	 *
1690 	 * [ This bug can be reproduced easily with a level-triggered
1691 	 *   PCI Ne2000 networking cards and PII/PIII processors, dual
1692 	 *   BX chipset. ]
1693 	 */
1694 	/*
1695 	 * Actually disabling the focus CPU check just makes the hang less
1696 	 * frequent as it makes the interrupt distributon model be more
1697 	 * like LRU than MRU (the short-term load is more even across CPUs).
1698 	 */
1699 
1700 	/*
1701 	 * - enable focus processor (bit==0)
1702 	 * - 64bit mode always use processor focus
1703 	 *   so no need to set it
1704 	 */
1705 	value &= ~APIC_SPIV_FOCUS_DISABLED;
1706 #endif
1707 
1708 	/*
1709 	 * Set spurious IRQ vector
1710 	 */
1711 	value |= SPURIOUS_APIC_VECTOR;
1712 	apic_write(APIC_SPIV, value);
1713 
1714 	perf_events_lapic_init();
1715 
1716 	/*
1717 	 * Set up LVT0, LVT1:
1718 	 *
1719 	 * set up through-local-APIC on the boot CPU's LINT0. This is not
1720 	 * strictly necessary in pure symmetric-IO mode, but sometimes
1721 	 * we delegate interrupts to the 8259A.
1722 	 */
1723 	/*
1724 	 * TODO: set up through-local-APIC from through-I/O-APIC? --macro
1725 	 */
1726 	value = apic_read(APIC_LVT0) & APIC_LVT_MASKED;
1727 	if (!cpu && (pic_mode || !value || skip_ioapic_setup)) {
1728 		value = APIC_DM_EXTINT;
1729 		apic_printk(APIC_VERBOSE, "enabled ExtINT on CPU#%d\n", cpu);
1730 	} else {
1731 		value = APIC_DM_EXTINT | APIC_LVT_MASKED;
1732 		apic_printk(APIC_VERBOSE, "masked ExtINT on CPU#%d\n", cpu);
1733 	}
1734 	apic_write(APIC_LVT0, value);
1735 
1736 	/*
1737 	 * Only the BSP sees the LINT1 NMI signal by default. This can be
1738 	 * modified by apic_extnmi= boot option.
1739 	 */
1740 	if ((!cpu && apic_extnmi != APIC_EXTNMI_NONE) ||
1741 	    apic_extnmi == APIC_EXTNMI_ALL)
1742 		value = APIC_DM_NMI;
1743 	else
1744 		value = APIC_DM_NMI | APIC_LVT_MASKED;
1745 
1746 	/* Is 82489DX ? */
1747 	if (!lapic_is_integrated())
1748 		value |= APIC_LVT_LEVEL_TRIGGER;
1749 	apic_write(APIC_LVT1, value);
1750 
1751 #ifdef CONFIG_X86_MCE_INTEL
1752 	/* Recheck CMCI information after local APIC is up on CPU #0 */
1753 	if (!cpu)
1754 		cmci_recheck();
1755 #endif
1756 }
1757 
1758 static void end_local_APIC_setup(void)
1759 {
1760 	lapic_setup_esr();
1761 
1762 #ifdef CONFIG_X86_32
1763 	{
1764 		unsigned int value;
1765 		/* Disable the local apic timer */
1766 		value = apic_read(APIC_LVTT);
1767 		value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
1768 		apic_write(APIC_LVTT, value);
1769 	}
1770 #endif
1771 
1772 	apic_pm_activate();
1773 }
1774 
1775 /*
1776  * APIC setup function for application processors. Called from smpboot.c
1777  */
1778 void apic_ap_setup(void)
1779 {
1780 	setup_local_APIC();
1781 	end_local_APIC_setup();
1782 }
1783 
1784 #ifdef CONFIG_X86_X2APIC
1785 int x2apic_mode;
1786 
1787 enum {
1788 	X2APIC_OFF,
1789 	X2APIC_ON,
1790 	X2APIC_DISABLED,
1791 };
1792 static int x2apic_state;
1793 
1794 static void __x2apic_disable(void)
1795 {
1796 	u64 msr;
1797 
1798 	if (!boot_cpu_has(X86_FEATURE_APIC))
1799 		return;
1800 
1801 	rdmsrl(MSR_IA32_APICBASE, msr);
1802 	if (!(msr & X2APIC_ENABLE))
1803 		return;
1804 	/* Disable xapic and x2apic first and then reenable xapic mode */
1805 	wrmsrl(MSR_IA32_APICBASE, msr & ~(X2APIC_ENABLE | XAPIC_ENABLE));
1806 	wrmsrl(MSR_IA32_APICBASE, msr & ~X2APIC_ENABLE);
1807 	printk_once(KERN_INFO "x2apic disabled\n");
1808 }
1809 
1810 static void __x2apic_enable(void)
1811 {
1812 	u64 msr;
1813 
1814 	rdmsrl(MSR_IA32_APICBASE, msr);
1815 	if (msr & X2APIC_ENABLE)
1816 		return;
1817 	wrmsrl(MSR_IA32_APICBASE, msr | X2APIC_ENABLE);
1818 	printk_once(KERN_INFO "x2apic enabled\n");
1819 }
1820 
1821 static int __init setup_nox2apic(char *str)
1822 {
1823 	if (x2apic_enabled()) {
1824 		int apicid = native_apic_msr_read(APIC_ID);
1825 
1826 		if (apicid >= 255) {
1827 			pr_warn("Apicid: %08x, cannot enforce nox2apic\n",
1828 				apicid);
1829 			return 0;
1830 		}
1831 		pr_warn("x2apic already enabled.\n");
1832 		__x2apic_disable();
1833 	}
1834 	setup_clear_cpu_cap(X86_FEATURE_X2APIC);
1835 	x2apic_state = X2APIC_DISABLED;
1836 	x2apic_mode = 0;
1837 	return 0;
1838 }
1839 early_param("nox2apic", setup_nox2apic);
1840 
1841 /* Called from cpu_init() to enable x2apic on (secondary) cpus */
1842 void x2apic_setup(void)
1843 {
1844 	/*
1845 	 * If x2apic is not in ON state, disable it if already enabled
1846 	 * from BIOS.
1847 	 */
1848 	if (x2apic_state != X2APIC_ON) {
1849 		__x2apic_disable();
1850 		return;
1851 	}
1852 	__x2apic_enable();
1853 }
1854 
1855 static __init void x2apic_disable(void)
1856 {
1857 	u32 x2apic_id, state = x2apic_state;
1858 
1859 	x2apic_mode = 0;
1860 	x2apic_state = X2APIC_DISABLED;
1861 
1862 	if (state != X2APIC_ON)
1863 		return;
1864 
1865 	x2apic_id = read_apic_id();
1866 	if (x2apic_id >= 255)
1867 		panic("Cannot disable x2apic, id: %08x\n", x2apic_id);
1868 
1869 	__x2apic_disable();
1870 	register_lapic_address(mp_lapic_addr);
1871 }
1872 
1873 static __init void x2apic_enable(void)
1874 {
1875 	if (x2apic_state != X2APIC_OFF)
1876 		return;
1877 
1878 	x2apic_mode = 1;
1879 	x2apic_state = X2APIC_ON;
1880 	__x2apic_enable();
1881 }
1882 
1883 static __init void try_to_enable_x2apic(int remap_mode)
1884 {
1885 	if (x2apic_state == X2APIC_DISABLED)
1886 		return;
1887 
1888 	if (remap_mode != IRQ_REMAP_X2APIC_MODE) {
1889 		/* IR is required if there is APIC ID > 255 even when running
1890 		 * under KVM
1891 		 */
1892 		if (max_physical_apicid > 255 ||
1893 		    !x86_init.hyper.x2apic_available()) {
1894 			pr_info("x2apic: IRQ remapping doesn't support X2APIC mode\n");
1895 			x2apic_disable();
1896 			return;
1897 		}
1898 
1899 		/*
1900 		 * without IR all CPUs can be addressed by IOAPIC/MSI
1901 		 * only in physical mode
1902 		 */
1903 		x2apic_phys = 1;
1904 	}
1905 	x2apic_enable();
1906 }
1907 
1908 void __init check_x2apic(void)
1909 {
1910 	if (x2apic_enabled()) {
1911 		pr_info("x2apic: enabled by BIOS, switching to x2apic ops\n");
1912 		x2apic_mode = 1;
1913 		x2apic_state = X2APIC_ON;
1914 	} else if (!boot_cpu_has(X86_FEATURE_X2APIC)) {
1915 		x2apic_state = X2APIC_DISABLED;
1916 	}
1917 }
1918 #else /* CONFIG_X86_X2APIC */
1919 static int __init validate_x2apic(void)
1920 {
1921 	if (!apic_is_x2apic_enabled())
1922 		return 0;
1923 	/*
1924 	 * Checkme: Can we simply turn off x2apic here instead of panic?
1925 	 */
1926 	panic("BIOS has enabled x2apic but kernel doesn't support x2apic, please disable x2apic in BIOS.\n");
1927 }
1928 early_initcall(validate_x2apic);
1929 
1930 static inline void try_to_enable_x2apic(int remap_mode) { }
1931 static inline void __x2apic_enable(void) { }
1932 #endif /* !CONFIG_X86_X2APIC */
1933 
1934 void __init enable_IR_x2apic(void)
1935 {
1936 	unsigned long flags;
1937 	int ret, ir_stat;
1938 
1939 	if (skip_ioapic_setup) {
1940 		pr_info("Not enabling interrupt remapping due to skipped IO-APIC setup\n");
1941 		return;
1942 	}
1943 
1944 	ir_stat = irq_remapping_prepare();
1945 	if (ir_stat < 0 && !x2apic_supported())
1946 		return;
1947 
1948 	ret = save_ioapic_entries();
1949 	if (ret) {
1950 		pr_info("Saving IO-APIC state failed: %d\n", ret);
1951 		return;
1952 	}
1953 
1954 	local_irq_save(flags);
1955 	legacy_pic->mask_all();
1956 	mask_ioapic_entries();
1957 
1958 	/* If irq_remapping_prepare() succeeded, try to enable it */
1959 	if (ir_stat >= 0)
1960 		ir_stat = irq_remapping_enable();
1961 	/* ir_stat contains the remap mode or an error code */
1962 	try_to_enable_x2apic(ir_stat);
1963 
1964 	if (ir_stat < 0)
1965 		restore_ioapic_entries();
1966 	legacy_pic->restore_mask();
1967 	local_irq_restore(flags);
1968 }
1969 
1970 #ifdef CONFIG_X86_64
1971 /*
1972  * Detect and enable local APICs on non-SMP boards.
1973  * Original code written by Keir Fraser.
1974  * On AMD64 we trust the BIOS - if it says no APIC it is likely
1975  * not correctly set up (usually the APIC timer won't work etc.)
1976  */
1977 static int __init detect_init_APIC(void)
1978 {
1979 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1980 		pr_info("No local APIC present\n");
1981 		return -1;
1982 	}
1983 
1984 	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1985 	return 0;
1986 }
1987 #else
1988 
1989 static int __init apic_verify(void)
1990 {
1991 	u32 features, h, l;
1992 
1993 	/*
1994 	 * The APIC feature bit should now be enabled
1995 	 * in `cpuid'
1996 	 */
1997 	features = cpuid_edx(1);
1998 	if (!(features & (1 << X86_FEATURE_APIC))) {
1999 		pr_warn("Could not enable APIC!\n");
2000 		return -1;
2001 	}
2002 	set_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
2003 	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
2004 
2005 	/* The BIOS may have set up the APIC at some other address */
2006 	if (boot_cpu_data.x86 >= 6) {
2007 		rdmsr(MSR_IA32_APICBASE, l, h);
2008 		if (l & MSR_IA32_APICBASE_ENABLE)
2009 			mp_lapic_addr = l & MSR_IA32_APICBASE_BASE;
2010 	}
2011 
2012 	pr_info("Found and enabled local APIC!\n");
2013 	return 0;
2014 }
2015 
2016 int __init apic_force_enable(unsigned long addr)
2017 {
2018 	u32 h, l;
2019 
2020 	if (disable_apic)
2021 		return -1;
2022 
2023 	/*
2024 	 * Some BIOSes disable the local APIC in the APIC_BASE
2025 	 * MSR. This can only be done in software for Intel P6 or later
2026 	 * and AMD K7 (Model > 1) or later.
2027 	 */
2028 	if (boot_cpu_data.x86 >= 6) {
2029 		rdmsr(MSR_IA32_APICBASE, l, h);
2030 		if (!(l & MSR_IA32_APICBASE_ENABLE)) {
2031 			pr_info("Local APIC disabled by BIOS -- reenabling.\n");
2032 			l &= ~MSR_IA32_APICBASE_BASE;
2033 			l |= MSR_IA32_APICBASE_ENABLE | addr;
2034 			wrmsr(MSR_IA32_APICBASE, l, h);
2035 			enabled_via_apicbase = 1;
2036 		}
2037 	}
2038 	return apic_verify();
2039 }
2040 
2041 /*
2042  * Detect and initialize APIC
2043  */
2044 static int __init detect_init_APIC(void)
2045 {
2046 	/* Disabled by kernel option? */
2047 	if (disable_apic)
2048 		return -1;
2049 
2050 	switch (boot_cpu_data.x86_vendor) {
2051 	case X86_VENDOR_AMD:
2052 		if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model > 1) ||
2053 		    (boot_cpu_data.x86 >= 15))
2054 			break;
2055 		goto no_apic;
2056 	case X86_VENDOR_HYGON:
2057 		break;
2058 	case X86_VENDOR_INTEL:
2059 		if (boot_cpu_data.x86 == 6 || boot_cpu_data.x86 == 15 ||
2060 		    (boot_cpu_data.x86 == 5 && boot_cpu_has(X86_FEATURE_APIC)))
2061 			break;
2062 		goto no_apic;
2063 	default:
2064 		goto no_apic;
2065 	}
2066 
2067 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
2068 		/*
2069 		 * Over-ride BIOS and try to enable the local APIC only if
2070 		 * "lapic" specified.
2071 		 */
2072 		if (!force_enable_local_apic) {
2073 			pr_info("Local APIC disabled by BIOS -- "
2074 				"you can enable it with \"lapic\"\n");
2075 			return -1;
2076 		}
2077 		if (apic_force_enable(APIC_DEFAULT_PHYS_BASE))
2078 			return -1;
2079 	} else {
2080 		if (apic_verify())
2081 			return -1;
2082 	}
2083 
2084 	apic_pm_activate();
2085 
2086 	return 0;
2087 
2088 no_apic:
2089 	pr_info("No local APIC present or hardware disabled\n");
2090 	return -1;
2091 }
2092 #endif
2093 
2094 /**
2095  * init_apic_mappings - initialize APIC mappings
2096  */
2097 void __init init_apic_mappings(void)
2098 {
2099 	unsigned int new_apicid;
2100 
2101 	apic_check_deadline_errata();
2102 
2103 	if (x2apic_mode) {
2104 		boot_cpu_physical_apicid = read_apic_id();
2105 		return;
2106 	}
2107 
2108 	/* If no local APIC can be found return early */
2109 	if (!smp_found_config && detect_init_APIC()) {
2110 		/* lets NOP'ify apic operations */
2111 		pr_info("APIC: disable apic facility\n");
2112 		apic_disable();
2113 	} else {
2114 		apic_phys = mp_lapic_addr;
2115 
2116 		/*
2117 		 * If the system has ACPI MADT tables or MP info, the LAPIC
2118 		 * address is already registered.
2119 		 */
2120 		if (!acpi_lapic && !smp_found_config)
2121 			register_lapic_address(apic_phys);
2122 	}
2123 
2124 	/*
2125 	 * Fetch the APIC ID of the BSP in case we have a
2126 	 * default configuration (or the MP table is broken).
2127 	 */
2128 	new_apicid = read_apic_id();
2129 	if (boot_cpu_physical_apicid != new_apicid) {
2130 		boot_cpu_physical_apicid = new_apicid;
2131 		/*
2132 		 * yeah -- we lie about apic_version
2133 		 * in case if apic was disabled via boot option
2134 		 * but it's not a problem for SMP compiled kernel
2135 		 * since apic_intr_mode_select is prepared for such
2136 		 * a case and disable smp mode
2137 		 */
2138 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
2139 	}
2140 }
2141 
2142 void __init register_lapic_address(unsigned long address)
2143 {
2144 	mp_lapic_addr = address;
2145 
2146 	if (!x2apic_mode) {
2147 		set_fixmap_nocache(FIX_APIC_BASE, address);
2148 		apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
2149 			    APIC_BASE, address);
2150 	}
2151 	if (boot_cpu_physical_apicid == -1U) {
2152 		boot_cpu_physical_apicid  = read_apic_id();
2153 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
2154 	}
2155 }
2156 
2157 /*
2158  * Local APIC interrupts
2159  */
2160 
2161 /*
2162  * This interrupt should _never_ happen with our APIC/SMP architecture
2163  */
2164 __visible void __irq_entry smp_spurious_interrupt(struct pt_regs *regs)
2165 {
2166 	u8 vector = ~regs->orig_ax;
2167 	u32 v;
2168 
2169 	entering_irq();
2170 	trace_spurious_apic_entry(vector);
2171 
2172 	inc_irq_stat(irq_spurious_count);
2173 
2174 	/*
2175 	 * If this is a spurious interrupt then do not acknowledge
2176 	 */
2177 	if (vector == SPURIOUS_APIC_VECTOR) {
2178 		/* See SDM vol 3 */
2179 		pr_info("Spurious APIC interrupt (vector 0xFF) on CPU#%d, should never happen.\n",
2180 			smp_processor_id());
2181 		goto out;
2182 	}
2183 
2184 	/*
2185 	 * If it is a vectored one, verify it's set in the ISR. If set,
2186 	 * acknowledge it.
2187 	 */
2188 	v = apic_read(APIC_ISR + ((vector & ~0x1f) >> 1));
2189 	if (v & (1 << (vector & 0x1f))) {
2190 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Acked\n",
2191 			vector, smp_processor_id());
2192 		ack_APIC_irq();
2193 	} else {
2194 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Not pending!\n",
2195 			vector, smp_processor_id());
2196 	}
2197 out:
2198 	trace_spurious_apic_exit(vector);
2199 	exiting_irq();
2200 }
2201 
2202 /*
2203  * This interrupt should never happen with our APIC/SMP architecture
2204  */
2205 __visible void __irq_entry smp_error_interrupt(struct pt_regs *regs)
2206 {
2207 	static const char * const error_interrupt_reason[] = {
2208 		"Send CS error",		/* APIC Error Bit 0 */
2209 		"Receive CS error",		/* APIC Error Bit 1 */
2210 		"Send accept error",		/* APIC Error Bit 2 */
2211 		"Receive accept error",		/* APIC Error Bit 3 */
2212 		"Redirectable IPI",		/* APIC Error Bit 4 */
2213 		"Send illegal vector",		/* APIC Error Bit 5 */
2214 		"Received illegal vector",	/* APIC Error Bit 6 */
2215 		"Illegal register address",	/* APIC Error Bit 7 */
2216 	};
2217 	u32 v, i = 0;
2218 
2219 	entering_irq();
2220 	trace_error_apic_entry(ERROR_APIC_VECTOR);
2221 
2222 	/* First tickle the hardware, only then report what went on. -- REW */
2223 	if (lapic_get_maxlvt() > 3)	/* Due to the Pentium erratum 3AP. */
2224 		apic_write(APIC_ESR, 0);
2225 	v = apic_read(APIC_ESR);
2226 	ack_APIC_irq();
2227 	atomic_inc(&irq_err_count);
2228 
2229 	apic_printk(APIC_DEBUG, KERN_DEBUG "APIC error on CPU%d: %02x",
2230 		    smp_processor_id(), v);
2231 
2232 	v &= 0xff;
2233 	while (v) {
2234 		if (v & 0x1)
2235 			apic_printk(APIC_DEBUG, KERN_CONT " : %s", error_interrupt_reason[i]);
2236 		i++;
2237 		v >>= 1;
2238 	}
2239 
2240 	apic_printk(APIC_DEBUG, KERN_CONT "\n");
2241 
2242 	trace_error_apic_exit(ERROR_APIC_VECTOR);
2243 	exiting_irq();
2244 }
2245 
2246 /**
2247  * connect_bsp_APIC - attach the APIC to the interrupt system
2248  */
2249 static void __init connect_bsp_APIC(void)
2250 {
2251 #ifdef CONFIG_X86_32
2252 	if (pic_mode) {
2253 		/*
2254 		 * Do not trust the local APIC being empty at bootup.
2255 		 */
2256 		clear_local_APIC();
2257 		/*
2258 		 * PIC mode, enable APIC mode in the IMCR, i.e.  connect BSP's
2259 		 * local APIC to INT and NMI lines.
2260 		 */
2261 		apic_printk(APIC_VERBOSE, "leaving PIC mode, "
2262 				"enabling APIC mode.\n");
2263 		imcr_pic_to_apic();
2264 	}
2265 #endif
2266 }
2267 
2268 /**
2269  * disconnect_bsp_APIC - detach the APIC from the interrupt system
2270  * @virt_wire_setup:	indicates, whether virtual wire mode is selected
2271  *
2272  * Virtual wire mode is necessary to deliver legacy interrupts even when the
2273  * APIC is disabled.
2274  */
2275 void disconnect_bsp_APIC(int virt_wire_setup)
2276 {
2277 	unsigned int value;
2278 
2279 #ifdef CONFIG_X86_32
2280 	if (pic_mode) {
2281 		/*
2282 		 * Put the board back into PIC mode (has an effect only on
2283 		 * certain older boards).  Note that APIC interrupts, including
2284 		 * IPIs, won't work beyond this point!  The only exception are
2285 		 * INIT IPIs.
2286 		 */
2287 		apic_printk(APIC_VERBOSE, "disabling APIC mode, "
2288 				"entering PIC mode.\n");
2289 		imcr_apic_to_pic();
2290 		return;
2291 	}
2292 #endif
2293 
2294 	/* Go back to Virtual Wire compatibility mode */
2295 
2296 	/* For the spurious interrupt use vector F, and enable it */
2297 	value = apic_read(APIC_SPIV);
2298 	value &= ~APIC_VECTOR_MASK;
2299 	value |= APIC_SPIV_APIC_ENABLED;
2300 	value |= 0xf;
2301 	apic_write(APIC_SPIV, value);
2302 
2303 	if (!virt_wire_setup) {
2304 		/*
2305 		 * For LVT0 make it edge triggered, active high,
2306 		 * external and enabled
2307 		 */
2308 		value = apic_read(APIC_LVT0);
2309 		value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2310 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2311 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2312 		value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2313 		value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_EXTINT);
2314 		apic_write(APIC_LVT0, value);
2315 	} else {
2316 		/* Disable LVT0 */
2317 		apic_write(APIC_LVT0, APIC_LVT_MASKED);
2318 	}
2319 
2320 	/*
2321 	 * For LVT1 make it edge triggered, active high,
2322 	 * nmi and enabled
2323 	 */
2324 	value = apic_read(APIC_LVT1);
2325 	value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2326 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2327 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2328 	value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2329 	value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_NMI);
2330 	apic_write(APIC_LVT1, value);
2331 }
2332 
2333 /*
2334  * The number of allocated logical CPU IDs. Since logical CPU IDs are allocated
2335  * contiguously, it equals to current allocated max logical CPU ID plus 1.
2336  * All allocated CPU IDs should be in the [0, nr_logical_cpuids) range,
2337  * so the maximum of nr_logical_cpuids is nr_cpu_ids.
2338  *
2339  * NOTE: Reserve 0 for BSP.
2340  */
2341 static int nr_logical_cpuids = 1;
2342 
2343 /*
2344  * Used to store mapping between logical CPU IDs and APIC IDs.
2345  */
2346 static int cpuid_to_apicid[] = {
2347 	[0 ... NR_CPUS - 1] = -1,
2348 };
2349 
2350 #ifdef CONFIG_SMP
2351 /**
2352  * apic_id_is_primary_thread - Check whether APIC ID belongs to a primary thread
2353  * @apicid: APIC ID to check
2354  */
2355 bool apic_id_is_primary_thread(unsigned int apicid)
2356 {
2357 	u32 mask;
2358 
2359 	if (smp_num_siblings == 1)
2360 		return true;
2361 	/* Isolate the SMT bit(s) in the APICID and check for 0 */
2362 	mask = (1U << (fls(smp_num_siblings) - 1)) - 1;
2363 	return !(apicid & mask);
2364 }
2365 #endif
2366 
2367 /*
2368  * Should use this API to allocate logical CPU IDs to keep nr_logical_cpuids
2369  * and cpuid_to_apicid[] synchronized.
2370  */
2371 static int allocate_logical_cpuid(int apicid)
2372 {
2373 	int i;
2374 
2375 	/*
2376 	 * cpuid <-> apicid mapping is persistent, so when a cpu is up,
2377 	 * check if the kernel has allocated a cpuid for it.
2378 	 */
2379 	for (i = 0; i < nr_logical_cpuids; i++) {
2380 		if (cpuid_to_apicid[i] == apicid)
2381 			return i;
2382 	}
2383 
2384 	/* Allocate a new cpuid. */
2385 	if (nr_logical_cpuids >= nr_cpu_ids) {
2386 		WARN_ONCE(1, "APIC: NR_CPUS/possible_cpus limit of %u reached. "
2387 			     "Processor %d/0x%x and the rest are ignored.\n",
2388 			     nr_cpu_ids, nr_logical_cpuids, apicid);
2389 		return -EINVAL;
2390 	}
2391 
2392 	cpuid_to_apicid[nr_logical_cpuids] = apicid;
2393 	return nr_logical_cpuids++;
2394 }
2395 
2396 int generic_processor_info(int apicid, int version)
2397 {
2398 	int cpu, max = nr_cpu_ids;
2399 	bool boot_cpu_detected = physid_isset(boot_cpu_physical_apicid,
2400 				phys_cpu_present_map);
2401 
2402 	/*
2403 	 * boot_cpu_physical_apicid is designed to have the apicid
2404 	 * returned by read_apic_id(), i.e, the apicid of the
2405 	 * currently booting-up processor. However, on some platforms,
2406 	 * it is temporarily modified by the apicid reported as BSP
2407 	 * through MP table. Concretely:
2408 	 *
2409 	 * - arch/x86/kernel/mpparse.c: MP_processor_info()
2410 	 * - arch/x86/mm/amdtopology.c: amd_numa_init()
2411 	 *
2412 	 * This function is executed with the modified
2413 	 * boot_cpu_physical_apicid. So, disabled_cpu_apicid kernel
2414 	 * parameter doesn't work to disable APs on kdump 2nd kernel.
2415 	 *
2416 	 * Since fixing handling of boot_cpu_physical_apicid requires
2417 	 * another discussion and tests on each platform, we leave it
2418 	 * for now and here we use read_apic_id() directly in this
2419 	 * function, generic_processor_info().
2420 	 */
2421 	if (disabled_cpu_apicid != BAD_APICID &&
2422 	    disabled_cpu_apicid != read_apic_id() &&
2423 	    disabled_cpu_apicid == apicid) {
2424 		int thiscpu = num_processors + disabled_cpus;
2425 
2426 		pr_warn("APIC: Disabling requested cpu."
2427 			" Processor %d/0x%x ignored.\n", thiscpu, apicid);
2428 
2429 		disabled_cpus++;
2430 		return -ENODEV;
2431 	}
2432 
2433 	/*
2434 	 * If boot cpu has not been detected yet, then only allow upto
2435 	 * nr_cpu_ids - 1 processors and keep one slot free for boot cpu
2436 	 */
2437 	if (!boot_cpu_detected && num_processors >= nr_cpu_ids - 1 &&
2438 	    apicid != boot_cpu_physical_apicid) {
2439 		int thiscpu = max + disabled_cpus - 1;
2440 
2441 		pr_warn("APIC: NR_CPUS/possible_cpus limit of %i almost"
2442 			" reached. Keeping one slot for boot cpu."
2443 			"  Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
2444 
2445 		disabled_cpus++;
2446 		return -ENODEV;
2447 	}
2448 
2449 	if (num_processors >= nr_cpu_ids) {
2450 		int thiscpu = max + disabled_cpus;
2451 
2452 		pr_warn("APIC: NR_CPUS/possible_cpus limit of %i reached. "
2453 			"Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
2454 
2455 		disabled_cpus++;
2456 		return -EINVAL;
2457 	}
2458 
2459 	if (apicid == boot_cpu_physical_apicid) {
2460 		/*
2461 		 * x86_bios_cpu_apicid is required to have processors listed
2462 		 * in same order as logical cpu numbers. Hence the first
2463 		 * entry is BSP, and so on.
2464 		 * boot_cpu_init() already hold bit 0 in cpu_present_mask
2465 		 * for BSP.
2466 		 */
2467 		cpu = 0;
2468 
2469 		/* Logical cpuid 0 is reserved for BSP. */
2470 		cpuid_to_apicid[0] = apicid;
2471 	} else {
2472 		cpu = allocate_logical_cpuid(apicid);
2473 		if (cpu < 0) {
2474 			disabled_cpus++;
2475 			return -EINVAL;
2476 		}
2477 	}
2478 
2479 	/*
2480 	 * Validate version
2481 	 */
2482 	if (version == 0x0) {
2483 		pr_warn("BIOS bug: APIC version is 0 for CPU %d/0x%x, fixing up to 0x10\n",
2484 			cpu, apicid);
2485 		version = 0x10;
2486 	}
2487 
2488 	if (version != boot_cpu_apic_version) {
2489 		pr_warn("BIOS bug: APIC version mismatch, boot CPU: %x, CPU %d: version %x\n",
2490 			boot_cpu_apic_version, cpu, version);
2491 	}
2492 
2493 	if (apicid > max_physical_apicid)
2494 		max_physical_apicid = apicid;
2495 
2496 #if defined(CONFIG_SMP) || defined(CONFIG_X86_64)
2497 	early_per_cpu(x86_cpu_to_apicid, cpu) = apicid;
2498 	early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
2499 #endif
2500 #ifdef CONFIG_X86_32
2501 	early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
2502 		apic->x86_32_early_logical_apicid(cpu);
2503 #endif
2504 	set_cpu_possible(cpu, true);
2505 	physid_set(apicid, phys_cpu_present_map);
2506 	set_cpu_present(cpu, true);
2507 	num_processors++;
2508 
2509 	return cpu;
2510 }
2511 
2512 int hard_smp_processor_id(void)
2513 {
2514 	return read_apic_id();
2515 }
2516 
2517 /*
2518  * Override the generic EOI implementation with an optimized version.
2519  * Only called during early boot when only one CPU is active and with
2520  * interrupts disabled, so we know this does not race with actual APIC driver
2521  * use.
2522  */
2523 void __init apic_set_eoi_write(void (*eoi_write)(u32 reg, u32 v))
2524 {
2525 	struct apic **drv;
2526 
2527 	for (drv = __apicdrivers; drv < __apicdrivers_end; drv++) {
2528 		/* Should happen once for each apic */
2529 		WARN_ON((*drv)->eoi_write == eoi_write);
2530 		(*drv)->native_eoi_write = (*drv)->eoi_write;
2531 		(*drv)->eoi_write = eoi_write;
2532 	}
2533 }
2534 
2535 static void __init apic_bsp_up_setup(void)
2536 {
2537 #ifdef CONFIG_X86_64
2538 	apic_write(APIC_ID, apic->set_apic_id(boot_cpu_physical_apicid));
2539 #else
2540 	/*
2541 	 * Hack: In case of kdump, after a crash, kernel might be booting
2542 	 * on a cpu with non-zero lapic id. But boot_cpu_physical_apicid
2543 	 * might be zero if read from MP tables. Get it from LAPIC.
2544 	 */
2545 # ifdef CONFIG_CRASH_DUMP
2546 	boot_cpu_physical_apicid = read_apic_id();
2547 # endif
2548 #endif
2549 	physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
2550 }
2551 
2552 /**
2553  * apic_bsp_setup - Setup function for local apic and io-apic
2554  * @upmode:		Force UP mode (for APIC_init_uniprocessor)
2555  */
2556 static void __init apic_bsp_setup(bool upmode)
2557 {
2558 	connect_bsp_APIC();
2559 	if (upmode)
2560 		apic_bsp_up_setup();
2561 	setup_local_APIC();
2562 
2563 	enable_IO_APIC();
2564 	end_local_APIC_setup();
2565 	irq_remap_enable_fault_handling();
2566 	setup_IO_APIC();
2567 }
2568 
2569 #ifdef CONFIG_UP_LATE_INIT
2570 void __init up_late_init(void)
2571 {
2572 	if (apic_intr_mode == APIC_PIC)
2573 		return;
2574 
2575 	/* Setup local timer */
2576 	x86_init.timers.setup_percpu_clockev();
2577 }
2578 #endif
2579 
2580 /*
2581  * Power management
2582  */
2583 #ifdef CONFIG_PM
2584 
2585 static struct {
2586 	/*
2587 	 * 'active' is true if the local APIC was enabled by us and
2588 	 * not the BIOS; this signifies that we are also responsible
2589 	 * for disabling it before entering apm/acpi suspend
2590 	 */
2591 	int active;
2592 	/* r/w apic fields */
2593 	unsigned int apic_id;
2594 	unsigned int apic_taskpri;
2595 	unsigned int apic_ldr;
2596 	unsigned int apic_dfr;
2597 	unsigned int apic_spiv;
2598 	unsigned int apic_lvtt;
2599 	unsigned int apic_lvtpc;
2600 	unsigned int apic_lvt0;
2601 	unsigned int apic_lvt1;
2602 	unsigned int apic_lvterr;
2603 	unsigned int apic_tmict;
2604 	unsigned int apic_tdcr;
2605 	unsigned int apic_thmr;
2606 	unsigned int apic_cmci;
2607 } apic_pm_state;
2608 
2609 static int lapic_suspend(void)
2610 {
2611 	unsigned long flags;
2612 	int maxlvt;
2613 
2614 	if (!apic_pm_state.active)
2615 		return 0;
2616 
2617 	maxlvt = lapic_get_maxlvt();
2618 
2619 	apic_pm_state.apic_id = apic_read(APIC_ID);
2620 	apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
2621 	apic_pm_state.apic_ldr = apic_read(APIC_LDR);
2622 	apic_pm_state.apic_dfr = apic_read(APIC_DFR);
2623 	apic_pm_state.apic_spiv = apic_read(APIC_SPIV);
2624 	apic_pm_state.apic_lvtt = apic_read(APIC_LVTT);
2625 	if (maxlvt >= 4)
2626 		apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC);
2627 	apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0);
2628 	apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1);
2629 	apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR);
2630 	apic_pm_state.apic_tmict = apic_read(APIC_TMICT);
2631 	apic_pm_state.apic_tdcr = apic_read(APIC_TDCR);
2632 #ifdef CONFIG_X86_THERMAL_VECTOR
2633 	if (maxlvt >= 5)
2634 		apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
2635 #endif
2636 #ifdef CONFIG_X86_MCE_INTEL
2637 	if (maxlvt >= 6)
2638 		apic_pm_state.apic_cmci = apic_read(APIC_LVTCMCI);
2639 #endif
2640 
2641 	local_irq_save(flags);
2642 
2643 	/*
2644 	 * Mask IOAPIC before disabling the local APIC to prevent stale IRR
2645 	 * entries on some implementations.
2646 	 */
2647 	mask_ioapic_entries();
2648 
2649 	disable_local_APIC();
2650 
2651 	irq_remapping_disable();
2652 
2653 	local_irq_restore(flags);
2654 	return 0;
2655 }
2656 
2657 static void lapic_resume(void)
2658 {
2659 	unsigned int l, h;
2660 	unsigned long flags;
2661 	int maxlvt;
2662 
2663 	if (!apic_pm_state.active)
2664 		return;
2665 
2666 	local_irq_save(flags);
2667 
2668 	/*
2669 	 * IO-APIC and PIC have their own resume routines.
2670 	 * We just mask them here to make sure the interrupt
2671 	 * subsystem is completely quiet while we enable x2apic
2672 	 * and interrupt-remapping.
2673 	 */
2674 	mask_ioapic_entries();
2675 	legacy_pic->mask_all();
2676 
2677 	if (x2apic_mode) {
2678 		__x2apic_enable();
2679 	} else {
2680 		/*
2681 		 * Make sure the APICBASE points to the right address
2682 		 *
2683 		 * FIXME! This will be wrong if we ever support suspend on
2684 		 * SMP! We'll need to do this as part of the CPU restore!
2685 		 */
2686 		if (boot_cpu_data.x86 >= 6) {
2687 			rdmsr(MSR_IA32_APICBASE, l, h);
2688 			l &= ~MSR_IA32_APICBASE_BASE;
2689 			l |= MSR_IA32_APICBASE_ENABLE | mp_lapic_addr;
2690 			wrmsr(MSR_IA32_APICBASE, l, h);
2691 		}
2692 	}
2693 
2694 	maxlvt = lapic_get_maxlvt();
2695 	apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED);
2696 	apic_write(APIC_ID, apic_pm_state.apic_id);
2697 	apic_write(APIC_DFR, apic_pm_state.apic_dfr);
2698 	apic_write(APIC_LDR, apic_pm_state.apic_ldr);
2699 	apic_write(APIC_TASKPRI, apic_pm_state.apic_taskpri);
2700 	apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
2701 	apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
2702 	apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
2703 #ifdef CONFIG_X86_THERMAL_VECTOR
2704 	if (maxlvt >= 5)
2705 		apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
2706 #endif
2707 #ifdef CONFIG_X86_MCE_INTEL
2708 	if (maxlvt >= 6)
2709 		apic_write(APIC_LVTCMCI, apic_pm_state.apic_cmci);
2710 #endif
2711 	if (maxlvt >= 4)
2712 		apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
2713 	apic_write(APIC_LVTT, apic_pm_state.apic_lvtt);
2714 	apic_write(APIC_TDCR, apic_pm_state.apic_tdcr);
2715 	apic_write(APIC_TMICT, apic_pm_state.apic_tmict);
2716 	apic_write(APIC_ESR, 0);
2717 	apic_read(APIC_ESR);
2718 	apic_write(APIC_LVTERR, apic_pm_state.apic_lvterr);
2719 	apic_write(APIC_ESR, 0);
2720 	apic_read(APIC_ESR);
2721 
2722 	irq_remapping_reenable(x2apic_mode);
2723 
2724 	local_irq_restore(flags);
2725 }
2726 
2727 /*
2728  * This device has no shutdown method - fully functioning local APICs
2729  * are needed on every CPU up until machine_halt/restart/poweroff.
2730  */
2731 
2732 static struct syscore_ops lapic_syscore_ops = {
2733 	.resume		= lapic_resume,
2734 	.suspend	= lapic_suspend,
2735 };
2736 
2737 static void apic_pm_activate(void)
2738 {
2739 	apic_pm_state.active = 1;
2740 }
2741 
2742 static int __init init_lapic_sysfs(void)
2743 {
2744 	/* XXX: remove suspend/resume procs if !apic_pm_state.active? */
2745 	if (boot_cpu_has(X86_FEATURE_APIC))
2746 		register_syscore_ops(&lapic_syscore_ops);
2747 
2748 	return 0;
2749 }
2750 
2751 /* local apic needs to resume before other devices access its registers. */
2752 core_initcall(init_lapic_sysfs);
2753 
2754 #else	/* CONFIG_PM */
2755 
2756 static void apic_pm_activate(void) { }
2757 
2758 #endif	/* CONFIG_PM */
2759 
2760 #ifdef CONFIG_X86_64
2761 
2762 static int multi_checked;
2763 static int multi;
2764 
2765 static int set_multi(const struct dmi_system_id *d)
2766 {
2767 	if (multi)
2768 		return 0;
2769 	pr_info("APIC: %s detected, Multi Chassis\n", d->ident);
2770 	multi = 1;
2771 	return 0;
2772 }
2773 
2774 static const struct dmi_system_id multi_dmi_table[] = {
2775 	{
2776 		.callback = set_multi,
2777 		.ident = "IBM System Summit2",
2778 		.matches = {
2779 			DMI_MATCH(DMI_SYS_VENDOR, "IBM"),
2780 			DMI_MATCH(DMI_PRODUCT_NAME, "Summit2"),
2781 		},
2782 	},
2783 	{}
2784 };
2785 
2786 static void dmi_check_multi(void)
2787 {
2788 	if (multi_checked)
2789 		return;
2790 
2791 	dmi_check_system(multi_dmi_table);
2792 	multi_checked = 1;
2793 }
2794 
2795 /*
2796  * apic_is_clustered_box() -- Check if we can expect good TSC
2797  *
2798  * Thus far, the major user of this is IBM's Summit2 series:
2799  * Clustered boxes may have unsynced TSC problems if they are
2800  * multi-chassis.
2801  * Use DMI to check them
2802  */
2803 int apic_is_clustered_box(void)
2804 {
2805 	dmi_check_multi();
2806 	return multi;
2807 }
2808 #endif
2809 
2810 /*
2811  * APIC command line parameters
2812  */
2813 static int __init setup_disableapic(char *arg)
2814 {
2815 	disable_apic = 1;
2816 	setup_clear_cpu_cap(X86_FEATURE_APIC);
2817 	return 0;
2818 }
2819 early_param("disableapic", setup_disableapic);
2820 
2821 /* same as disableapic, for compatibility */
2822 static int __init setup_nolapic(char *arg)
2823 {
2824 	return setup_disableapic(arg);
2825 }
2826 early_param("nolapic", setup_nolapic);
2827 
2828 static int __init parse_lapic_timer_c2_ok(char *arg)
2829 {
2830 	local_apic_timer_c2_ok = 1;
2831 	return 0;
2832 }
2833 early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok);
2834 
2835 static int __init parse_disable_apic_timer(char *arg)
2836 {
2837 	disable_apic_timer = 1;
2838 	return 0;
2839 }
2840 early_param("noapictimer", parse_disable_apic_timer);
2841 
2842 static int __init parse_nolapic_timer(char *arg)
2843 {
2844 	disable_apic_timer = 1;
2845 	return 0;
2846 }
2847 early_param("nolapic_timer", parse_nolapic_timer);
2848 
2849 static int __init apic_set_verbosity(char *arg)
2850 {
2851 	if (!arg)  {
2852 #ifdef CONFIG_X86_64
2853 		skip_ioapic_setup = 0;
2854 		return 0;
2855 #endif
2856 		return -EINVAL;
2857 	}
2858 
2859 	if (strcmp("debug", arg) == 0)
2860 		apic_verbosity = APIC_DEBUG;
2861 	else if (strcmp("verbose", arg) == 0)
2862 		apic_verbosity = APIC_VERBOSE;
2863 #ifdef CONFIG_X86_64
2864 	else {
2865 		pr_warn("APIC Verbosity level %s not recognised"
2866 			" use apic=verbose or apic=debug\n", arg);
2867 		return -EINVAL;
2868 	}
2869 #endif
2870 
2871 	return 0;
2872 }
2873 early_param("apic", apic_set_verbosity);
2874 
2875 static int __init lapic_insert_resource(void)
2876 {
2877 	if (!apic_phys)
2878 		return -1;
2879 
2880 	/* Put local APIC into the resource map. */
2881 	lapic_resource.start = apic_phys;
2882 	lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
2883 	insert_resource(&iomem_resource, &lapic_resource);
2884 
2885 	return 0;
2886 }
2887 
2888 /*
2889  * need call insert after e820__reserve_resources()
2890  * that is using request_resource
2891  */
2892 late_initcall(lapic_insert_resource);
2893 
2894 static int __init apic_set_disabled_cpu_apicid(char *arg)
2895 {
2896 	if (!arg || !get_option(&arg, &disabled_cpu_apicid))
2897 		return -EINVAL;
2898 
2899 	return 0;
2900 }
2901 early_param("disable_cpu_apicid", apic_set_disabled_cpu_apicid);
2902 
2903 static int __init apic_set_extnmi(char *arg)
2904 {
2905 	if (!arg)
2906 		return -EINVAL;
2907 
2908 	if (!strncmp("all", arg, 3))
2909 		apic_extnmi = APIC_EXTNMI_ALL;
2910 	else if (!strncmp("none", arg, 4))
2911 		apic_extnmi = APIC_EXTNMI_NONE;
2912 	else if (!strncmp("bsp", arg, 3))
2913 		apic_extnmi = APIC_EXTNMI_BSP;
2914 	else {
2915 		pr_warn("Unknown external NMI delivery mode `%s' ignored\n", arg);
2916 		return -EINVAL;
2917 	}
2918 
2919 	return 0;
2920 }
2921 early_param("apic_extnmi", apic_set_extnmi);
2922