xref: /linux/arch/x86/kernel/apic/apic.c (revision f474808acb3c4b30552d9c59b181244e0300d218)
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 = -1U;
69 EXPORT_SYMBOL_GPL(boot_cpu_physical_apicid);
70 
71 u8 boot_cpu_apic_version;
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 __read_mostly = 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 = 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;
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;
176 int disable_apic;
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;
181 EXPORT_SYMBOL_GPL(local_apic_timer_c2_ok);
182 
183 /*
184  * Debug level, exported for io_apic.c
185  */
186 int apic_verbosity;
187 
188 int pic_mode;
189 
190 /* Have we found an MP table */
191 int smp_found_config;
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;
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_XEON_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_CORE,	0x22),
597 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_HASWELL_ULT,	0x20),
598 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_HASWELL_GT3E,	0x17),
599 
600 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_BROADWELL_CORE,	0x25),
601 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_BROADWELL_GT3E,	0x17),
602 
603 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_SKYLAKE_MOBILE,	0xb2),
604 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_SKYLAKE_DESKTOP,	0xb2),
605 
606 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_KABYLAKE_MOBILE,	0x52),
607 	DEADLINE_MODEL_MATCH_REV ( INTEL_FAM6_KABYLAKE_DESKTOP,	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.
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_warning("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? */
834 	if (!boot_cpu_has(X86_FEATURE_APIC))
835 		return true;
836 
837 	/* Deadline timer is based on TSC so no further PIT action required */
838 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
839 		return false;
840 
841 	/* APIC timer disabled? */
842 	if (disable_apic_timer)
843 		return true;
844 	/*
845 	 * The APIC timer frequency is known already, no PIT calibration
846 	 * required. If unknown, let the PIT be initialized.
847 	 */
848 	return lapic_timer_period == 0;
849 }
850 
851 static int __init calibrate_APIC_clock(void)
852 {
853 	struct clock_event_device *levt = this_cpu_ptr(&lapic_events);
854 	void (*real_handler)(struct clock_event_device *dev);
855 	unsigned long deltaj;
856 	long delta, deltatsc;
857 	int pm_referenced = 0;
858 
859 	if (boot_cpu_has(X86_FEATURE_TSC_DEADLINE_TIMER))
860 		return 0;
861 
862 	/*
863 	 * Check if lapic timer has already been calibrated by platform
864 	 * specific routine, such as tsc calibration code. If so just fill
865 	 * in the clockevent structure and return.
866 	 */
867 	if (!lapic_init_clockevent()) {
868 		apic_printk(APIC_VERBOSE, "lapic timer already calibrated %d\n",
869 			    lapic_timer_period);
870 		/*
871 		 * Direct calibration methods must have an always running
872 		 * local APIC timer, no need for broadcast timer.
873 		 */
874 		lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
875 		return 0;
876 	}
877 
878 	apic_printk(APIC_VERBOSE, "Using local APIC timer interrupts.\n"
879 		    "calibrating APIC timer ...\n");
880 
881 	local_irq_disable();
882 
883 	/* Replace the global interrupt handler */
884 	real_handler = global_clock_event->event_handler;
885 	global_clock_event->event_handler = lapic_cal_handler;
886 
887 	/*
888 	 * Setup the APIC counter to maximum. There is no way the lapic
889 	 * can underflow in the 100ms detection time frame
890 	 */
891 	__setup_APIC_LVTT(0xffffffff, 0, 0);
892 
893 	/* Let the interrupts run */
894 	local_irq_enable();
895 
896 	while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
897 		cpu_relax();
898 
899 	local_irq_disable();
900 
901 	/* Restore the real event handler */
902 	global_clock_event->event_handler = real_handler;
903 
904 	/* Build delta t1-t2 as apic timer counts down */
905 	delta = lapic_cal_t1 - lapic_cal_t2;
906 	apic_printk(APIC_VERBOSE, "... lapic delta = %ld\n", delta);
907 
908 	deltatsc = (long)(lapic_cal_tsc2 - lapic_cal_tsc1);
909 
910 	/* we trust the PM based calibration if possible */
911 	pm_referenced = !calibrate_by_pmtimer(lapic_cal_pm2 - lapic_cal_pm1,
912 					&delta, &deltatsc);
913 
914 	lapic_timer_period = (delta * APIC_DIVISOR) / LAPIC_CAL_LOOPS;
915 	lapic_init_clockevent();
916 
917 	apic_printk(APIC_VERBOSE, "..... delta %ld\n", delta);
918 	apic_printk(APIC_VERBOSE, "..... mult: %u\n", lapic_clockevent.mult);
919 	apic_printk(APIC_VERBOSE, "..... calibration result: %u\n",
920 		    lapic_timer_period);
921 
922 	if (boot_cpu_has(X86_FEATURE_TSC)) {
923 		apic_printk(APIC_VERBOSE, "..... CPU clock speed is "
924 			    "%ld.%04ld MHz.\n",
925 			    (deltatsc / LAPIC_CAL_LOOPS) / (1000000 / HZ),
926 			    (deltatsc / LAPIC_CAL_LOOPS) % (1000000 / HZ));
927 	}
928 
929 	apic_printk(APIC_VERBOSE, "..... host bus clock speed is "
930 		    "%u.%04u MHz.\n",
931 		    lapic_timer_period / (1000000 / HZ),
932 		    lapic_timer_period % (1000000 / HZ));
933 
934 	/*
935 	 * Do a sanity check on the APIC calibration result
936 	 */
937 	if (lapic_timer_period < (1000000 / HZ)) {
938 		local_irq_enable();
939 		pr_warning("APIC frequency too slow, disabling apic timer\n");
940 		return -1;
941 	}
942 
943 	levt->features &= ~CLOCK_EVT_FEAT_DUMMY;
944 
945 	/*
946 	 * PM timer calibration failed or not turned on
947 	 * so lets try APIC timer based calibration
948 	 */
949 	if (!pm_referenced) {
950 		apic_printk(APIC_VERBOSE, "... verify APIC timer\n");
951 
952 		/*
953 		 * Setup the apic timer manually
954 		 */
955 		levt->event_handler = lapic_cal_handler;
956 		lapic_timer_set_periodic(levt);
957 		lapic_cal_loops = -1;
958 
959 		/* Let the interrupts run */
960 		local_irq_enable();
961 
962 		while (lapic_cal_loops <= LAPIC_CAL_LOOPS)
963 			cpu_relax();
964 
965 		/* Stop the lapic timer */
966 		local_irq_disable();
967 		lapic_timer_shutdown(levt);
968 
969 		/* Jiffies delta */
970 		deltaj = lapic_cal_j2 - lapic_cal_j1;
971 		apic_printk(APIC_VERBOSE, "... jiffies delta = %lu\n", deltaj);
972 
973 		/* Check, if the jiffies result is consistent */
974 		if (deltaj >= LAPIC_CAL_LOOPS-2 && deltaj <= LAPIC_CAL_LOOPS+2)
975 			apic_printk(APIC_VERBOSE, "... jiffies result ok\n");
976 		else
977 			levt->features |= CLOCK_EVT_FEAT_DUMMY;
978 	}
979 	local_irq_enable();
980 
981 	if (levt->features & CLOCK_EVT_FEAT_DUMMY) {
982 		pr_warning("APIC timer disabled due to verification failure\n");
983 		return -1;
984 	}
985 
986 	return 0;
987 }
988 
989 /*
990  * Setup the boot APIC
991  *
992  * Calibrate and verify the result.
993  */
994 void __init setup_boot_APIC_clock(void)
995 {
996 	/*
997 	 * The local apic timer can be disabled via the kernel
998 	 * commandline or from the CPU detection code. Register the lapic
999 	 * timer as a dummy clock event source on SMP systems, so the
1000 	 * broadcast mechanism is used. On UP systems simply ignore it.
1001 	 */
1002 	if (disable_apic_timer) {
1003 		pr_info("Disabling APIC timer\n");
1004 		/* No broadcast on UP ! */
1005 		if (num_possible_cpus() > 1) {
1006 			lapic_clockevent.mult = 1;
1007 			setup_APIC_timer();
1008 		}
1009 		return;
1010 	}
1011 
1012 	if (calibrate_APIC_clock()) {
1013 		/* No broadcast on UP ! */
1014 		if (num_possible_cpus() > 1)
1015 			setup_APIC_timer();
1016 		return;
1017 	}
1018 
1019 	/*
1020 	 * If nmi_watchdog is set to IO_APIC, we need the
1021 	 * PIT/HPET going.  Otherwise register lapic as a dummy
1022 	 * device.
1023 	 */
1024 	lapic_clockevent.features &= ~CLOCK_EVT_FEAT_DUMMY;
1025 
1026 	/* Setup the lapic or request the broadcast */
1027 	setup_APIC_timer();
1028 	amd_e400_c1e_apic_setup();
1029 }
1030 
1031 void setup_secondary_APIC_clock(void)
1032 {
1033 	setup_APIC_timer();
1034 	amd_e400_c1e_apic_setup();
1035 }
1036 
1037 /*
1038  * The guts of the apic timer interrupt
1039  */
1040 static void local_apic_timer_interrupt(void)
1041 {
1042 	struct clock_event_device *evt = this_cpu_ptr(&lapic_events);
1043 
1044 	/*
1045 	 * Normally we should not be here till LAPIC has been initialized but
1046 	 * in some cases like kdump, its possible that there is a pending LAPIC
1047 	 * timer interrupt from previous kernel's context and is delivered in
1048 	 * new kernel the moment interrupts are enabled.
1049 	 *
1050 	 * Interrupts are enabled early and LAPIC is setup much later, hence
1051 	 * its possible that when we get here evt->event_handler is NULL.
1052 	 * Check for event_handler being NULL and discard the interrupt as
1053 	 * spurious.
1054 	 */
1055 	if (!evt->event_handler) {
1056 		pr_warning("Spurious LAPIC timer interrupt on cpu %d\n",
1057 			   smp_processor_id());
1058 		/* Switch it off */
1059 		lapic_timer_shutdown(evt);
1060 		return;
1061 	}
1062 
1063 	/*
1064 	 * the NMI deadlock-detector uses this.
1065 	 */
1066 	inc_irq_stat(apic_timer_irqs);
1067 
1068 	evt->event_handler(evt);
1069 }
1070 
1071 /*
1072  * Local APIC timer interrupt. This is the most natural way for doing
1073  * local interrupts, but local timer interrupts can be emulated by
1074  * broadcast interrupts too. [in case the hw doesn't support APIC timers]
1075  *
1076  * [ if a single-CPU system runs an SMP kernel then we call the local
1077  *   interrupt as well. Thus we cannot inline the local irq ... ]
1078  */
1079 __visible void __irq_entry smp_apic_timer_interrupt(struct pt_regs *regs)
1080 {
1081 	struct pt_regs *old_regs = set_irq_regs(regs);
1082 
1083 	/*
1084 	 * NOTE! We'd better ACK the irq immediately,
1085 	 * because timer handling can be slow.
1086 	 *
1087 	 * update_process_times() expects us to have done irq_enter().
1088 	 * Besides, if we don't timer interrupts ignore the global
1089 	 * interrupt lock, which is the WrongThing (tm) to do.
1090 	 */
1091 	entering_ack_irq();
1092 	trace_local_timer_entry(LOCAL_TIMER_VECTOR);
1093 	local_apic_timer_interrupt();
1094 	trace_local_timer_exit(LOCAL_TIMER_VECTOR);
1095 	exiting_irq();
1096 
1097 	set_irq_regs(old_regs);
1098 }
1099 
1100 int setup_profiling_timer(unsigned int multiplier)
1101 {
1102 	return -EINVAL;
1103 }
1104 
1105 /*
1106  * Local APIC start and shutdown
1107  */
1108 
1109 /**
1110  * clear_local_APIC - shutdown the local APIC
1111  *
1112  * This is called, when a CPU is disabled and before rebooting, so the state of
1113  * the local APIC has no dangling leftovers. Also used to cleanout any BIOS
1114  * leftovers during boot.
1115  */
1116 void clear_local_APIC(void)
1117 {
1118 	int maxlvt;
1119 	u32 v;
1120 
1121 	/* APIC hasn't been mapped yet */
1122 	if (!x2apic_mode && !apic_phys)
1123 		return;
1124 
1125 	maxlvt = lapic_get_maxlvt();
1126 	/*
1127 	 * Masking an LVT entry can trigger a local APIC error
1128 	 * if the vector is zero. Mask LVTERR first to prevent this.
1129 	 */
1130 	if (maxlvt >= 3) {
1131 		v = ERROR_APIC_VECTOR; /* any non-zero vector will do */
1132 		apic_write(APIC_LVTERR, v | APIC_LVT_MASKED);
1133 	}
1134 	/*
1135 	 * Careful: we have to set masks only first to deassert
1136 	 * any level-triggered sources.
1137 	 */
1138 	v = apic_read(APIC_LVTT);
1139 	apic_write(APIC_LVTT, v | APIC_LVT_MASKED);
1140 	v = apic_read(APIC_LVT0);
1141 	apic_write(APIC_LVT0, v | APIC_LVT_MASKED);
1142 	v = apic_read(APIC_LVT1);
1143 	apic_write(APIC_LVT1, v | APIC_LVT_MASKED);
1144 	if (maxlvt >= 4) {
1145 		v = apic_read(APIC_LVTPC);
1146 		apic_write(APIC_LVTPC, v | APIC_LVT_MASKED);
1147 	}
1148 
1149 	/* lets not touch this if we didn't frob it */
1150 #ifdef CONFIG_X86_THERMAL_VECTOR
1151 	if (maxlvt >= 5) {
1152 		v = apic_read(APIC_LVTTHMR);
1153 		apic_write(APIC_LVTTHMR, v | APIC_LVT_MASKED);
1154 	}
1155 #endif
1156 #ifdef CONFIG_X86_MCE_INTEL
1157 	if (maxlvt >= 6) {
1158 		v = apic_read(APIC_LVTCMCI);
1159 		if (!(v & APIC_LVT_MASKED))
1160 			apic_write(APIC_LVTCMCI, v | APIC_LVT_MASKED);
1161 	}
1162 #endif
1163 
1164 	/*
1165 	 * Clean APIC state for other OSs:
1166 	 */
1167 	apic_write(APIC_LVTT, APIC_LVT_MASKED);
1168 	apic_write(APIC_LVT0, APIC_LVT_MASKED);
1169 	apic_write(APIC_LVT1, APIC_LVT_MASKED);
1170 	if (maxlvt >= 3)
1171 		apic_write(APIC_LVTERR, APIC_LVT_MASKED);
1172 	if (maxlvt >= 4)
1173 		apic_write(APIC_LVTPC, APIC_LVT_MASKED);
1174 
1175 	/* Integrated APIC (!82489DX) ? */
1176 	if (lapic_is_integrated()) {
1177 		if (maxlvt > 3)
1178 			/* Clear ESR due to Pentium errata 3AP and 11AP */
1179 			apic_write(APIC_ESR, 0);
1180 		apic_read(APIC_ESR);
1181 	}
1182 }
1183 
1184 /**
1185  * disable_local_APIC - clear and disable the local APIC
1186  */
1187 void disable_local_APIC(void)
1188 {
1189 	unsigned int value;
1190 
1191 	/* APIC hasn't been mapped yet */
1192 	if (!x2apic_mode && !apic_phys)
1193 		return;
1194 
1195 	clear_local_APIC();
1196 
1197 	/*
1198 	 * Disable APIC (implies clearing of registers
1199 	 * for 82489DX!).
1200 	 */
1201 	value = apic_read(APIC_SPIV);
1202 	value &= ~APIC_SPIV_APIC_ENABLED;
1203 	apic_write(APIC_SPIV, value);
1204 
1205 #ifdef CONFIG_X86_32
1206 	/*
1207 	 * When LAPIC was disabled by the BIOS and enabled by the kernel,
1208 	 * restore the disabled state.
1209 	 */
1210 	if (enabled_via_apicbase) {
1211 		unsigned int l, h;
1212 
1213 		rdmsr(MSR_IA32_APICBASE, l, h);
1214 		l &= ~MSR_IA32_APICBASE_ENABLE;
1215 		wrmsr(MSR_IA32_APICBASE, l, h);
1216 	}
1217 #endif
1218 }
1219 
1220 /*
1221  * If Linux enabled the LAPIC against the BIOS default disable it down before
1222  * re-entering the BIOS on shutdown.  Otherwise the BIOS may get confused and
1223  * not power-off.  Additionally clear all LVT entries before disable_local_APIC
1224  * for the case where Linux didn't enable the LAPIC.
1225  */
1226 void lapic_shutdown(void)
1227 {
1228 	unsigned long flags;
1229 
1230 	if (!boot_cpu_has(X86_FEATURE_APIC) && !apic_from_smp_config())
1231 		return;
1232 
1233 	local_irq_save(flags);
1234 
1235 #ifdef CONFIG_X86_32
1236 	if (!enabled_via_apicbase)
1237 		clear_local_APIC();
1238 	else
1239 #endif
1240 		disable_local_APIC();
1241 
1242 
1243 	local_irq_restore(flags);
1244 }
1245 
1246 /**
1247  * sync_Arb_IDs - synchronize APIC bus arbitration IDs
1248  */
1249 void __init sync_Arb_IDs(void)
1250 {
1251 	/*
1252 	 * Unsupported on P4 - see Intel Dev. Manual Vol. 3, Ch. 8.6.1 And not
1253 	 * needed on AMD.
1254 	 */
1255 	if (modern_apic() || boot_cpu_data.x86_vendor == X86_VENDOR_AMD)
1256 		return;
1257 
1258 	/*
1259 	 * Wait for idle.
1260 	 */
1261 	apic_wait_icr_idle();
1262 
1263 	apic_printk(APIC_DEBUG, "Synchronizing Arb IDs.\n");
1264 	apic_write(APIC_ICR, APIC_DEST_ALLINC |
1265 			APIC_INT_LEVELTRIG | APIC_DM_INIT);
1266 }
1267 
1268 enum apic_intr_mode_id apic_intr_mode;
1269 
1270 static int __init apic_intr_mode_select(void)
1271 {
1272 	/* Check kernel option */
1273 	if (disable_apic) {
1274 		pr_info("APIC disabled via kernel command line\n");
1275 		return APIC_PIC;
1276 	}
1277 
1278 	/* Check BIOS */
1279 #ifdef CONFIG_X86_64
1280 	/* On 64-bit, the APIC must be integrated, Check local APIC only */
1281 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1282 		disable_apic = 1;
1283 		pr_info("APIC disabled by BIOS\n");
1284 		return APIC_PIC;
1285 	}
1286 #else
1287 	/* On 32-bit, the APIC may be integrated APIC or 82489DX */
1288 
1289 	/* Neither 82489DX nor integrated APIC ? */
1290 	if (!boot_cpu_has(X86_FEATURE_APIC) && !smp_found_config) {
1291 		disable_apic = 1;
1292 		return APIC_PIC;
1293 	}
1294 
1295 	/* If the BIOS pretends there is an integrated APIC ? */
1296 	if (!boot_cpu_has(X86_FEATURE_APIC) &&
1297 		APIC_INTEGRATED(boot_cpu_apic_version)) {
1298 		disable_apic = 1;
1299 		pr_err(FW_BUG "Local APIC %d not detected, force emulation\n",
1300 				       boot_cpu_physical_apicid);
1301 		return APIC_PIC;
1302 	}
1303 #endif
1304 
1305 	/* Check MP table or ACPI MADT configuration */
1306 	if (!smp_found_config) {
1307 		disable_ioapic_support();
1308 		if (!acpi_lapic) {
1309 			pr_info("APIC: ACPI MADT or MP tables are not detected\n");
1310 			return APIC_VIRTUAL_WIRE_NO_CONFIG;
1311 		}
1312 		return APIC_VIRTUAL_WIRE;
1313 	}
1314 
1315 #ifdef CONFIG_SMP
1316 	/* If SMP should be disabled, then really disable it! */
1317 	if (!setup_max_cpus) {
1318 		pr_info("APIC: SMP mode deactivated\n");
1319 		return APIC_SYMMETRIC_IO_NO_ROUTING;
1320 	}
1321 
1322 	if (read_apic_id() != boot_cpu_physical_apicid) {
1323 		panic("Boot APIC ID in local APIC unexpected (%d vs %d)",
1324 		     read_apic_id(), boot_cpu_physical_apicid);
1325 		/* Or can we switch back to PIC here? */
1326 	}
1327 #endif
1328 
1329 	return APIC_SYMMETRIC_IO;
1330 }
1331 
1332 /*
1333  * An initial setup of the virtual wire mode.
1334  */
1335 void __init init_bsp_APIC(void)
1336 {
1337 	unsigned int value;
1338 
1339 	/*
1340 	 * Don't do the setup now if we have a SMP BIOS as the
1341 	 * through-I/O-APIC virtual wire mode might be active.
1342 	 */
1343 	if (smp_found_config || !boot_cpu_has(X86_FEATURE_APIC))
1344 		return;
1345 
1346 	/*
1347 	 * Do not trust the local APIC being empty at bootup.
1348 	 */
1349 	clear_local_APIC();
1350 
1351 	/*
1352 	 * Enable APIC.
1353 	 */
1354 	value = apic_read(APIC_SPIV);
1355 	value &= ~APIC_VECTOR_MASK;
1356 	value |= APIC_SPIV_APIC_ENABLED;
1357 
1358 #ifdef CONFIG_X86_32
1359 	/* This bit is reserved on P4/Xeon and should be cleared */
1360 	if ((boot_cpu_data.x86_vendor == X86_VENDOR_INTEL) &&
1361 	    (boot_cpu_data.x86 == 15))
1362 		value &= ~APIC_SPIV_FOCUS_DISABLED;
1363 	else
1364 #endif
1365 		value |= APIC_SPIV_FOCUS_DISABLED;
1366 	value |= SPURIOUS_APIC_VECTOR;
1367 	apic_write(APIC_SPIV, value);
1368 
1369 	/*
1370 	 * Set up the virtual wire mode.
1371 	 */
1372 	apic_write(APIC_LVT0, APIC_DM_EXTINT);
1373 	value = APIC_DM_NMI;
1374 	if (!lapic_is_integrated())		/* 82489DX */
1375 		value |= APIC_LVT_LEVEL_TRIGGER;
1376 	if (apic_extnmi == APIC_EXTNMI_NONE)
1377 		value |= APIC_LVT_MASKED;
1378 	apic_write(APIC_LVT1, value);
1379 }
1380 
1381 static void __init apic_bsp_setup(bool upmode);
1382 
1383 /* Init the interrupt delivery mode for the BSP */
1384 void __init apic_intr_mode_init(void)
1385 {
1386 	bool upmode = IS_ENABLED(CONFIG_UP_LATE_INIT);
1387 
1388 	apic_intr_mode = apic_intr_mode_select();
1389 
1390 	switch (apic_intr_mode) {
1391 	case APIC_PIC:
1392 		pr_info("APIC: Keep in PIC mode(8259)\n");
1393 		return;
1394 	case APIC_VIRTUAL_WIRE:
1395 		pr_info("APIC: Switch to virtual wire mode setup\n");
1396 		default_setup_apic_routing();
1397 		break;
1398 	case APIC_VIRTUAL_WIRE_NO_CONFIG:
1399 		pr_info("APIC: Switch to virtual wire mode setup with no configuration\n");
1400 		upmode = true;
1401 		default_setup_apic_routing();
1402 		break;
1403 	case APIC_SYMMETRIC_IO:
1404 		pr_info("APIC: Switch to symmetric I/O mode setup\n");
1405 		default_setup_apic_routing();
1406 		break;
1407 	case APIC_SYMMETRIC_IO_NO_ROUTING:
1408 		pr_info("APIC: Switch to symmetric I/O mode setup in no SMP routine\n");
1409 		break;
1410 	}
1411 
1412 	apic_bsp_setup(upmode);
1413 }
1414 
1415 static void lapic_setup_esr(void)
1416 {
1417 	unsigned int oldvalue, value, maxlvt;
1418 
1419 	if (!lapic_is_integrated()) {
1420 		pr_info("No ESR for 82489DX.\n");
1421 		return;
1422 	}
1423 
1424 	if (apic->disable_esr) {
1425 		/*
1426 		 * Something untraceable is creating bad interrupts on
1427 		 * secondary quads ... for the moment, just leave the
1428 		 * ESR disabled - we can't do anything useful with the
1429 		 * errors anyway - mbligh
1430 		 */
1431 		pr_info("Leaving ESR disabled.\n");
1432 		return;
1433 	}
1434 
1435 	maxlvt = lapic_get_maxlvt();
1436 	if (maxlvt > 3)		/* Due to the Pentium erratum 3AP. */
1437 		apic_write(APIC_ESR, 0);
1438 	oldvalue = apic_read(APIC_ESR);
1439 
1440 	/* enables sending errors */
1441 	value = ERROR_APIC_VECTOR;
1442 	apic_write(APIC_LVTERR, value);
1443 
1444 	/*
1445 	 * spec says clear errors after enabling vector.
1446 	 */
1447 	if (maxlvt > 3)
1448 		apic_write(APIC_ESR, 0);
1449 	value = apic_read(APIC_ESR);
1450 	if (value != oldvalue)
1451 		apic_printk(APIC_VERBOSE, "ESR value before enabling "
1452 			"vector: 0x%08x  after: 0x%08x\n",
1453 			oldvalue, value);
1454 }
1455 
1456 static void apic_pending_intr_clear(void)
1457 {
1458 	long long max_loops = cpu_khz ? cpu_khz : 1000000;
1459 	unsigned long long tsc = 0, ntsc;
1460 	unsigned int queued;
1461 	unsigned long value;
1462 	int i, j, acked = 0;
1463 
1464 	if (boot_cpu_has(X86_FEATURE_TSC))
1465 		tsc = rdtsc();
1466 	/*
1467 	 * After a crash, we no longer service the interrupts and a pending
1468 	 * interrupt from previous kernel might still have ISR bit set.
1469 	 *
1470 	 * Most probably by now CPU has serviced that pending interrupt and
1471 	 * it might not have done the ack_APIC_irq() because it thought,
1472 	 * interrupt came from i8259 as ExtInt. LAPIC did not get EOI so it
1473 	 * does not clear the ISR bit and cpu thinks it has already serivced
1474 	 * the interrupt. Hence a vector might get locked. It was noticed
1475 	 * for timer irq (vector 0x31). Issue an extra EOI to clear ISR.
1476 	 */
1477 	do {
1478 		queued = 0;
1479 		for (i = APIC_ISR_NR - 1; i >= 0; i--)
1480 			queued |= apic_read(APIC_IRR + i*0x10);
1481 
1482 		for (i = APIC_ISR_NR - 1; i >= 0; i--) {
1483 			value = apic_read(APIC_ISR + i*0x10);
1484 			for_each_set_bit(j, &value, 32) {
1485 				ack_APIC_irq();
1486 				acked++;
1487 			}
1488 		}
1489 		if (acked > 256) {
1490 			pr_err("LAPIC pending interrupts after %d EOI\n", acked);
1491 			break;
1492 		}
1493 		if (queued) {
1494 			if (boot_cpu_has(X86_FEATURE_TSC) && cpu_khz) {
1495 				ntsc = rdtsc();
1496 				max_loops = (long long)cpu_khz << 10;
1497 				max_loops -= ntsc - tsc;
1498 			} else {
1499 				max_loops--;
1500 			}
1501 		}
1502 	} while (queued && max_loops > 0);
1503 	WARN_ON(max_loops <= 0);
1504 }
1505 
1506 /**
1507  * setup_local_APIC - setup the local APIC
1508  *
1509  * Used to setup local APIC while initializing BSP or bringing up APs.
1510  * Always called with preemption disabled.
1511  */
1512 static void setup_local_APIC(void)
1513 {
1514 	int cpu = smp_processor_id();
1515 	unsigned int value;
1516 #ifdef CONFIG_X86_32
1517 	int logical_apicid, ldr_apicid;
1518 #endif
1519 
1520 
1521 	if (disable_apic) {
1522 		disable_ioapic_support();
1523 		return;
1524 	}
1525 
1526 #ifdef CONFIG_X86_32
1527 	/* Pound the ESR really hard over the head with a big hammer - mbligh */
1528 	if (lapic_is_integrated() && apic->disable_esr) {
1529 		apic_write(APIC_ESR, 0);
1530 		apic_write(APIC_ESR, 0);
1531 		apic_write(APIC_ESR, 0);
1532 		apic_write(APIC_ESR, 0);
1533 	}
1534 #endif
1535 	perf_events_lapic_init();
1536 
1537 	/*
1538 	 * Double-check whether this APIC is really registered.
1539 	 * This is meaningless in clustered apic mode, so we skip it.
1540 	 */
1541 	BUG_ON(!apic->apic_id_registered());
1542 
1543 	/*
1544 	 * Intel recommends to set DFR, LDR and TPR before enabling
1545 	 * an APIC.  See e.g. "AP-388 82489DX User's Manual" (Intel
1546 	 * document number 292116).  So here it goes...
1547 	 */
1548 	apic->init_apic_ldr();
1549 
1550 #ifdef CONFIG_X86_32
1551 	/*
1552 	 * APIC LDR is initialized.  If logical_apicid mapping was
1553 	 * initialized during get_smp_config(), make sure it matches the
1554 	 * actual value.
1555 	 */
1556 	logical_apicid = early_per_cpu(x86_cpu_to_logical_apicid, cpu);
1557 	ldr_apicid = GET_APIC_LOGICAL_ID(apic_read(APIC_LDR));
1558 	WARN_ON(logical_apicid != BAD_APICID && logical_apicid != ldr_apicid);
1559 	/* always use the value from LDR */
1560 	early_per_cpu(x86_cpu_to_logical_apicid, cpu) = ldr_apicid;
1561 #endif
1562 
1563 	/*
1564 	 * Set Task Priority to 'accept all'. We never change this
1565 	 * later on.
1566 	 */
1567 	value = apic_read(APIC_TASKPRI);
1568 	value &= ~APIC_TPRI_MASK;
1569 	apic_write(APIC_TASKPRI, value);
1570 
1571 	apic_pending_intr_clear();
1572 
1573 	/*
1574 	 * Now that we are all set up, enable the APIC
1575 	 */
1576 	value = apic_read(APIC_SPIV);
1577 	value &= ~APIC_VECTOR_MASK;
1578 	/*
1579 	 * Enable APIC
1580 	 */
1581 	value |= APIC_SPIV_APIC_ENABLED;
1582 
1583 #ifdef CONFIG_X86_32
1584 	/*
1585 	 * Some unknown Intel IO/APIC (or APIC) errata is biting us with
1586 	 * certain networking cards. If high frequency interrupts are
1587 	 * happening on a particular IOAPIC pin, plus the IOAPIC routing
1588 	 * entry is masked/unmasked at a high rate as well then sooner or
1589 	 * later IOAPIC line gets 'stuck', no more interrupts are received
1590 	 * from the device. If focus CPU is disabled then the hang goes
1591 	 * away, oh well :-(
1592 	 *
1593 	 * [ This bug can be reproduced easily with a level-triggered
1594 	 *   PCI Ne2000 networking cards and PII/PIII processors, dual
1595 	 *   BX chipset. ]
1596 	 */
1597 	/*
1598 	 * Actually disabling the focus CPU check just makes the hang less
1599 	 * frequent as it makes the interrupt distributon model be more
1600 	 * like LRU than MRU (the short-term load is more even across CPUs).
1601 	 */
1602 
1603 	/*
1604 	 * - enable focus processor (bit==0)
1605 	 * - 64bit mode always use processor focus
1606 	 *   so no need to set it
1607 	 */
1608 	value &= ~APIC_SPIV_FOCUS_DISABLED;
1609 #endif
1610 
1611 	/*
1612 	 * Set spurious IRQ vector
1613 	 */
1614 	value |= SPURIOUS_APIC_VECTOR;
1615 	apic_write(APIC_SPIV, value);
1616 
1617 	/*
1618 	 * Set up LVT0, LVT1:
1619 	 *
1620 	 * set up through-local-APIC on the boot CPU's LINT0. This is not
1621 	 * strictly necessary in pure symmetric-IO mode, but sometimes
1622 	 * we delegate interrupts to the 8259A.
1623 	 */
1624 	/*
1625 	 * TODO: set up through-local-APIC from through-I/O-APIC? --macro
1626 	 */
1627 	value = apic_read(APIC_LVT0) & APIC_LVT_MASKED;
1628 	if (!cpu && (pic_mode || !value || skip_ioapic_setup)) {
1629 		value = APIC_DM_EXTINT;
1630 		apic_printk(APIC_VERBOSE, "enabled ExtINT on CPU#%d\n", cpu);
1631 	} else {
1632 		value = APIC_DM_EXTINT | APIC_LVT_MASKED;
1633 		apic_printk(APIC_VERBOSE, "masked ExtINT on CPU#%d\n", cpu);
1634 	}
1635 	apic_write(APIC_LVT0, value);
1636 
1637 	/*
1638 	 * Only the BSP sees the LINT1 NMI signal by default. This can be
1639 	 * modified by apic_extnmi= boot option.
1640 	 */
1641 	if ((!cpu && apic_extnmi != APIC_EXTNMI_NONE) ||
1642 	    apic_extnmi == APIC_EXTNMI_ALL)
1643 		value = APIC_DM_NMI;
1644 	else
1645 		value = APIC_DM_NMI | APIC_LVT_MASKED;
1646 
1647 	/* Is 82489DX ? */
1648 	if (!lapic_is_integrated())
1649 		value |= APIC_LVT_LEVEL_TRIGGER;
1650 	apic_write(APIC_LVT1, value);
1651 
1652 #ifdef CONFIG_X86_MCE_INTEL
1653 	/* Recheck CMCI information after local APIC is up on CPU #0 */
1654 	if (!cpu)
1655 		cmci_recheck();
1656 #endif
1657 }
1658 
1659 static void end_local_APIC_setup(void)
1660 {
1661 	lapic_setup_esr();
1662 
1663 #ifdef CONFIG_X86_32
1664 	{
1665 		unsigned int value;
1666 		/* Disable the local apic timer */
1667 		value = apic_read(APIC_LVTT);
1668 		value |= (APIC_LVT_MASKED | LOCAL_TIMER_VECTOR);
1669 		apic_write(APIC_LVTT, value);
1670 	}
1671 #endif
1672 
1673 	apic_pm_activate();
1674 }
1675 
1676 /*
1677  * APIC setup function for application processors. Called from smpboot.c
1678  */
1679 void apic_ap_setup(void)
1680 {
1681 	setup_local_APIC();
1682 	end_local_APIC_setup();
1683 }
1684 
1685 #ifdef CONFIG_X86_X2APIC
1686 int x2apic_mode;
1687 
1688 enum {
1689 	X2APIC_OFF,
1690 	X2APIC_ON,
1691 	X2APIC_DISABLED,
1692 };
1693 static int x2apic_state;
1694 
1695 static void __x2apic_disable(void)
1696 {
1697 	u64 msr;
1698 
1699 	if (!boot_cpu_has(X86_FEATURE_APIC))
1700 		return;
1701 
1702 	rdmsrl(MSR_IA32_APICBASE, msr);
1703 	if (!(msr & X2APIC_ENABLE))
1704 		return;
1705 	/* Disable xapic and x2apic first and then reenable xapic mode */
1706 	wrmsrl(MSR_IA32_APICBASE, msr & ~(X2APIC_ENABLE | XAPIC_ENABLE));
1707 	wrmsrl(MSR_IA32_APICBASE, msr & ~X2APIC_ENABLE);
1708 	printk_once(KERN_INFO "x2apic disabled\n");
1709 }
1710 
1711 static void __x2apic_enable(void)
1712 {
1713 	u64 msr;
1714 
1715 	rdmsrl(MSR_IA32_APICBASE, msr);
1716 	if (msr & X2APIC_ENABLE)
1717 		return;
1718 	wrmsrl(MSR_IA32_APICBASE, msr | X2APIC_ENABLE);
1719 	printk_once(KERN_INFO "x2apic enabled\n");
1720 }
1721 
1722 static int __init setup_nox2apic(char *str)
1723 {
1724 	if (x2apic_enabled()) {
1725 		int apicid = native_apic_msr_read(APIC_ID);
1726 
1727 		if (apicid >= 255) {
1728 			pr_warning("Apicid: %08x, cannot enforce nox2apic\n",
1729 				   apicid);
1730 			return 0;
1731 		}
1732 		pr_warning("x2apic already enabled.\n");
1733 		__x2apic_disable();
1734 	}
1735 	setup_clear_cpu_cap(X86_FEATURE_X2APIC);
1736 	x2apic_state = X2APIC_DISABLED;
1737 	x2apic_mode = 0;
1738 	return 0;
1739 }
1740 early_param("nox2apic", setup_nox2apic);
1741 
1742 /* Called from cpu_init() to enable x2apic on (secondary) cpus */
1743 void x2apic_setup(void)
1744 {
1745 	/*
1746 	 * If x2apic is not in ON state, disable it if already enabled
1747 	 * from BIOS.
1748 	 */
1749 	if (x2apic_state != X2APIC_ON) {
1750 		__x2apic_disable();
1751 		return;
1752 	}
1753 	__x2apic_enable();
1754 }
1755 
1756 static __init void x2apic_disable(void)
1757 {
1758 	u32 x2apic_id, state = x2apic_state;
1759 
1760 	x2apic_mode = 0;
1761 	x2apic_state = X2APIC_DISABLED;
1762 
1763 	if (state != X2APIC_ON)
1764 		return;
1765 
1766 	x2apic_id = read_apic_id();
1767 	if (x2apic_id >= 255)
1768 		panic("Cannot disable x2apic, id: %08x\n", x2apic_id);
1769 
1770 	__x2apic_disable();
1771 	register_lapic_address(mp_lapic_addr);
1772 }
1773 
1774 static __init void x2apic_enable(void)
1775 {
1776 	if (x2apic_state != X2APIC_OFF)
1777 		return;
1778 
1779 	x2apic_mode = 1;
1780 	x2apic_state = X2APIC_ON;
1781 	__x2apic_enable();
1782 }
1783 
1784 static __init void try_to_enable_x2apic(int remap_mode)
1785 {
1786 	if (x2apic_state == X2APIC_DISABLED)
1787 		return;
1788 
1789 	if (remap_mode != IRQ_REMAP_X2APIC_MODE) {
1790 		/* IR is required if there is APIC ID > 255 even when running
1791 		 * under KVM
1792 		 */
1793 		if (max_physical_apicid > 255 ||
1794 		    !x86_init.hyper.x2apic_available()) {
1795 			pr_info("x2apic: IRQ remapping doesn't support X2APIC mode\n");
1796 			x2apic_disable();
1797 			return;
1798 		}
1799 
1800 		/*
1801 		 * without IR all CPUs can be addressed by IOAPIC/MSI
1802 		 * only in physical mode
1803 		 */
1804 		x2apic_phys = 1;
1805 	}
1806 	x2apic_enable();
1807 }
1808 
1809 void __init check_x2apic(void)
1810 {
1811 	if (x2apic_enabled()) {
1812 		pr_info("x2apic: enabled by BIOS, switching to x2apic ops\n");
1813 		x2apic_mode = 1;
1814 		x2apic_state = X2APIC_ON;
1815 	} else if (!boot_cpu_has(X86_FEATURE_X2APIC)) {
1816 		x2apic_state = X2APIC_DISABLED;
1817 	}
1818 }
1819 #else /* CONFIG_X86_X2APIC */
1820 static int __init validate_x2apic(void)
1821 {
1822 	if (!apic_is_x2apic_enabled())
1823 		return 0;
1824 	/*
1825 	 * Checkme: Can we simply turn off x2apic here instead of panic?
1826 	 */
1827 	panic("BIOS has enabled x2apic but kernel doesn't support x2apic, please disable x2apic in BIOS.\n");
1828 }
1829 early_initcall(validate_x2apic);
1830 
1831 static inline void try_to_enable_x2apic(int remap_mode) { }
1832 static inline void __x2apic_enable(void) { }
1833 #endif /* !CONFIG_X86_X2APIC */
1834 
1835 void __init enable_IR_x2apic(void)
1836 {
1837 	unsigned long flags;
1838 	int ret, ir_stat;
1839 
1840 	if (skip_ioapic_setup) {
1841 		pr_info("Not enabling interrupt remapping due to skipped IO-APIC setup\n");
1842 		return;
1843 	}
1844 
1845 	ir_stat = irq_remapping_prepare();
1846 	if (ir_stat < 0 && !x2apic_supported())
1847 		return;
1848 
1849 	ret = save_ioapic_entries();
1850 	if (ret) {
1851 		pr_info("Saving IO-APIC state failed: %d\n", ret);
1852 		return;
1853 	}
1854 
1855 	local_irq_save(flags);
1856 	legacy_pic->mask_all();
1857 	mask_ioapic_entries();
1858 
1859 	/* If irq_remapping_prepare() succeeded, try to enable it */
1860 	if (ir_stat >= 0)
1861 		ir_stat = irq_remapping_enable();
1862 	/* ir_stat contains the remap mode or an error code */
1863 	try_to_enable_x2apic(ir_stat);
1864 
1865 	if (ir_stat < 0)
1866 		restore_ioapic_entries();
1867 	legacy_pic->restore_mask();
1868 	local_irq_restore(flags);
1869 }
1870 
1871 #ifdef CONFIG_X86_64
1872 /*
1873  * Detect and enable local APICs on non-SMP boards.
1874  * Original code written by Keir Fraser.
1875  * On AMD64 we trust the BIOS - if it says no APIC it is likely
1876  * not correctly set up (usually the APIC timer won't work etc.)
1877  */
1878 static int __init detect_init_APIC(void)
1879 {
1880 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1881 		pr_info("No local APIC present\n");
1882 		return -1;
1883 	}
1884 
1885 	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1886 	return 0;
1887 }
1888 #else
1889 
1890 static int __init apic_verify(void)
1891 {
1892 	u32 features, h, l;
1893 
1894 	/*
1895 	 * The APIC feature bit should now be enabled
1896 	 * in `cpuid'
1897 	 */
1898 	features = cpuid_edx(1);
1899 	if (!(features & (1 << X86_FEATURE_APIC))) {
1900 		pr_warning("Could not enable APIC!\n");
1901 		return -1;
1902 	}
1903 	set_cpu_cap(&boot_cpu_data, X86_FEATURE_APIC);
1904 	mp_lapic_addr = APIC_DEFAULT_PHYS_BASE;
1905 
1906 	/* The BIOS may have set up the APIC at some other address */
1907 	if (boot_cpu_data.x86 >= 6) {
1908 		rdmsr(MSR_IA32_APICBASE, l, h);
1909 		if (l & MSR_IA32_APICBASE_ENABLE)
1910 			mp_lapic_addr = l & MSR_IA32_APICBASE_BASE;
1911 	}
1912 
1913 	pr_info("Found and enabled local APIC!\n");
1914 	return 0;
1915 }
1916 
1917 int __init apic_force_enable(unsigned long addr)
1918 {
1919 	u32 h, l;
1920 
1921 	if (disable_apic)
1922 		return -1;
1923 
1924 	/*
1925 	 * Some BIOSes disable the local APIC in the APIC_BASE
1926 	 * MSR. This can only be done in software for Intel P6 or later
1927 	 * and AMD K7 (Model > 1) or later.
1928 	 */
1929 	if (boot_cpu_data.x86 >= 6) {
1930 		rdmsr(MSR_IA32_APICBASE, l, h);
1931 		if (!(l & MSR_IA32_APICBASE_ENABLE)) {
1932 			pr_info("Local APIC disabled by BIOS -- reenabling.\n");
1933 			l &= ~MSR_IA32_APICBASE_BASE;
1934 			l |= MSR_IA32_APICBASE_ENABLE | addr;
1935 			wrmsr(MSR_IA32_APICBASE, l, h);
1936 			enabled_via_apicbase = 1;
1937 		}
1938 	}
1939 	return apic_verify();
1940 }
1941 
1942 /*
1943  * Detect and initialize APIC
1944  */
1945 static int __init detect_init_APIC(void)
1946 {
1947 	/* Disabled by kernel option? */
1948 	if (disable_apic)
1949 		return -1;
1950 
1951 	switch (boot_cpu_data.x86_vendor) {
1952 	case X86_VENDOR_AMD:
1953 		if ((boot_cpu_data.x86 == 6 && boot_cpu_data.x86_model > 1) ||
1954 		    (boot_cpu_data.x86 >= 15))
1955 			break;
1956 		goto no_apic;
1957 	case X86_VENDOR_HYGON:
1958 		break;
1959 	case X86_VENDOR_INTEL:
1960 		if (boot_cpu_data.x86 == 6 || boot_cpu_data.x86 == 15 ||
1961 		    (boot_cpu_data.x86 == 5 && boot_cpu_has(X86_FEATURE_APIC)))
1962 			break;
1963 		goto no_apic;
1964 	default:
1965 		goto no_apic;
1966 	}
1967 
1968 	if (!boot_cpu_has(X86_FEATURE_APIC)) {
1969 		/*
1970 		 * Over-ride BIOS and try to enable the local APIC only if
1971 		 * "lapic" specified.
1972 		 */
1973 		if (!force_enable_local_apic) {
1974 			pr_info("Local APIC disabled by BIOS -- "
1975 				"you can enable it with \"lapic\"\n");
1976 			return -1;
1977 		}
1978 		if (apic_force_enable(APIC_DEFAULT_PHYS_BASE))
1979 			return -1;
1980 	} else {
1981 		if (apic_verify())
1982 			return -1;
1983 	}
1984 
1985 	apic_pm_activate();
1986 
1987 	return 0;
1988 
1989 no_apic:
1990 	pr_info("No local APIC present or hardware disabled\n");
1991 	return -1;
1992 }
1993 #endif
1994 
1995 /**
1996  * init_apic_mappings - initialize APIC mappings
1997  */
1998 void __init init_apic_mappings(void)
1999 {
2000 	unsigned int new_apicid;
2001 
2002 	apic_check_deadline_errata();
2003 
2004 	if (x2apic_mode) {
2005 		boot_cpu_physical_apicid = read_apic_id();
2006 		return;
2007 	}
2008 
2009 	/* If no local APIC can be found return early */
2010 	if (!smp_found_config && detect_init_APIC()) {
2011 		/* lets NOP'ify apic operations */
2012 		pr_info("APIC: disable apic facility\n");
2013 		apic_disable();
2014 	} else {
2015 		apic_phys = mp_lapic_addr;
2016 
2017 		/*
2018 		 * If the system has ACPI MADT tables or MP info, the LAPIC
2019 		 * address is already registered.
2020 		 */
2021 		if (!acpi_lapic && !smp_found_config)
2022 			register_lapic_address(apic_phys);
2023 	}
2024 
2025 	/*
2026 	 * Fetch the APIC ID of the BSP in case we have a
2027 	 * default configuration (or the MP table is broken).
2028 	 */
2029 	new_apicid = read_apic_id();
2030 	if (boot_cpu_physical_apicid != new_apicid) {
2031 		boot_cpu_physical_apicid = new_apicid;
2032 		/*
2033 		 * yeah -- we lie about apic_version
2034 		 * in case if apic was disabled via boot option
2035 		 * but it's not a problem for SMP compiled kernel
2036 		 * since apic_intr_mode_select is prepared for such
2037 		 * a case and disable smp mode
2038 		 */
2039 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
2040 	}
2041 }
2042 
2043 void __init register_lapic_address(unsigned long address)
2044 {
2045 	mp_lapic_addr = address;
2046 
2047 	if (!x2apic_mode) {
2048 		set_fixmap_nocache(FIX_APIC_BASE, address);
2049 		apic_printk(APIC_VERBOSE, "mapped APIC to %16lx (%16lx)\n",
2050 			    APIC_BASE, address);
2051 	}
2052 	if (boot_cpu_physical_apicid == -1U) {
2053 		boot_cpu_physical_apicid  = read_apic_id();
2054 		boot_cpu_apic_version = GET_APIC_VERSION(apic_read(APIC_LVR));
2055 	}
2056 }
2057 
2058 /*
2059  * Local APIC interrupts
2060  */
2061 
2062 /*
2063  * This interrupt should _never_ happen with our APIC/SMP architecture
2064  */
2065 __visible void __irq_entry smp_spurious_interrupt(struct pt_regs *regs)
2066 {
2067 	u8 vector = ~regs->orig_ax;
2068 	u32 v;
2069 
2070 	entering_irq();
2071 	trace_spurious_apic_entry(vector);
2072 
2073 	inc_irq_stat(irq_spurious_count);
2074 
2075 	/*
2076 	 * If this is a spurious interrupt then do not acknowledge
2077 	 */
2078 	if (vector == SPURIOUS_APIC_VECTOR) {
2079 		/* See SDM vol 3 */
2080 		pr_info("Spurious APIC interrupt (vector 0xFF) on CPU#%d, should never happen.\n",
2081 			smp_processor_id());
2082 		goto out;
2083 	}
2084 
2085 	/*
2086 	 * If it is a vectored one, verify it's set in the ISR. If set,
2087 	 * acknowledge it.
2088 	 */
2089 	v = apic_read(APIC_ISR + ((vector & ~0x1f) >> 1));
2090 	if (v & (1 << (vector & 0x1f))) {
2091 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Acked\n",
2092 			vector, smp_processor_id());
2093 		ack_APIC_irq();
2094 	} else {
2095 		pr_info("Spurious interrupt (vector 0x%02x) on CPU#%d. Not pending!\n",
2096 			vector, smp_processor_id());
2097 	}
2098 out:
2099 	trace_spurious_apic_exit(vector);
2100 	exiting_irq();
2101 }
2102 
2103 /*
2104  * This interrupt should never happen with our APIC/SMP architecture
2105  */
2106 __visible void __irq_entry smp_error_interrupt(struct pt_regs *regs)
2107 {
2108 	static const char * const error_interrupt_reason[] = {
2109 		"Send CS error",		/* APIC Error Bit 0 */
2110 		"Receive CS error",		/* APIC Error Bit 1 */
2111 		"Send accept error",		/* APIC Error Bit 2 */
2112 		"Receive accept error",		/* APIC Error Bit 3 */
2113 		"Redirectable IPI",		/* APIC Error Bit 4 */
2114 		"Send illegal vector",		/* APIC Error Bit 5 */
2115 		"Received illegal vector",	/* APIC Error Bit 6 */
2116 		"Illegal register address",	/* APIC Error Bit 7 */
2117 	};
2118 	u32 v, i = 0;
2119 
2120 	entering_irq();
2121 	trace_error_apic_entry(ERROR_APIC_VECTOR);
2122 
2123 	/* First tickle the hardware, only then report what went on. -- REW */
2124 	if (lapic_get_maxlvt() > 3)	/* Due to the Pentium erratum 3AP. */
2125 		apic_write(APIC_ESR, 0);
2126 	v = apic_read(APIC_ESR);
2127 	ack_APIC_irq();
2128 	atomic_inc(&irq_err_count);
2129 
2130 	apic_printk(APIC_DEBUG, KERN_DEBUG "APIC error on CPU%d: %02x",
2131 		    smp_processor_id(), v);
2132 
2133 	v &= 0xff;
2134 	while (v) {
2135 		if (v & 0x1)
2136 			apic_printk(APIC_DEBUG, KERN_CONT " : %s", error_interrupt_reason[i]);
2137 		i++;
2138 		v >>= 1;
2139 	}
2140 
2141 	apic_printk(APIC_DEBUG, KERN_CONT "\n");
2142 
2143 	trace_error_apic_exit(ERROR_APIC_VECTOR);
2144 	exiting_irq();
2145 }
2146 
2147 /**
2148  * connect_bsp_APIC - attach the APIC to the interrupt system
2149  */
2150 static void __init connect_bsp_APIC(void)
2151 {
2152 #ifdef CONFIG_X86_32
2153 	if (pic_mode) {
2154 		/*
2155 		 * Do not trust the local APIC being empty at bootup.
2156 		 */
2157 		clear_local_APIC();
2158 		/*
2159 		 * PIC mode, enable APIC mode in the IMCR, i.e.  connect BSP's
2160 		 * local APIC to INT and NMI lines.
2161 		 */
2162 		apic_printk(APIC_VERBOSE, "leaving PIC mode, "
2163 				"enabling APIC mode.\n");
2164 		imcr_pic_to_apic();
2165 	}
2166 #endif
2167 }
2168 
2169 /**
2170  * disconnect_bsp_APIC - detach the APIC from the interrupt system
2171  * @virt_wire_setup:	indicates, whether virtual wire mode is selected
2172  *
2173  * Virtual wire mode is necessary to deliver legacy interrupts even when the
2174  * APIC is disabled.
2175  */
2176 void disconnect_bsp_APIC(int virt_wire_setup)
2177 {
2178 	unsigned int value;
2179 
2180 #ifdef CONFIG_X86_32
2181 	if (pic_mode) {
2182 		/*
2183 		 * Put the board back into PIC mode (has an effect only on
2184 		 * certain older boards).  Note that APIC interrupts, including
2185 		 * IPIs, won't work beyond this point!  The only exception are
2186 		 * INIT IPIs.
2187 		 */
2188 		apic_printk(APIC_VERBOSE, "disabling APIC mode, "
2189 				"entering PIC mode.\n");
2190 		imcr_apic_to_pic();
2191 		return;
2192 	}
2193 #endif
2194 
2195 	/* Go back to Virtual Wire compatibility mode */
2196 
2197 	/* For the spurious interrupt use vector F, and enable it */
2198 	value = apic_read(APIC_SPIV);
2199 	value &= ~APIC_VECTOR_MASK;
2200 	value |= APIC_SPIV_APIC_ENABLED;
2201 	value |= 0xf;
2202 	apic_write(APIC_SPIV, value);
2203 
2204 	if (!virt_wire_setup) {
2205 		/*
2206 		 * For LVT0 make it edge triggered, active high,
2207 		 * external and enabled
2208 		 */
2209 		value = apic_read(APIC_LVT0);
2210 		value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2211 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2212 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2213 		value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2214 		value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_EXTINT);
2215 		apic_write(APIC_LVT0, value);
2216 	} else {
2217 		/* Disable LVT0 */
2218 		apic_write(APIC_LVT0, APIC_LVT_MASKED);
2219 	}
2220 
2221 	/*
2222 	 * For LVT1 make it edge triggered, active high,
2223 	 * nmi and enabled
2224 	 */
2225 	value = apic_read(APIC_LVT1);
2226 	value &= ~(APIC_MODE_MASK | APIC_SEND_PENDING |
2227 			APIC_INPUT_POLARITY | APIC_LVT_REMOTE_IRR |
2228 			APIC_LVT_LEVEL_TRIGGER | APIC_LVT_MASKED);
2229 	value |= APIC_LVT_REMOTE_IRR | APIC_SEND_PENDING;
2230 	value = SET_APIC_DELIVERY_MODE(value, APIC_MODE_NMI);
2231 	apic_write(APIC_LVT1, value);
2232 }
2233 
2234 /*
2235  * The number of allocated logical CPU IDs. Since logical CPU IDs are allocated
2236  * contiguously, it equals to current allocated max logical CPU ID plus 1.
2237  * All allocated CPU IDs should be in the [0, nr_logical_cpuids) range,
2238  * so the maximum of nr_logical_cpuids is nr_cpu_ids.
2239  *
2240  * NOTE: Reserve 0 for BSP.
2241  */
2242 static int nr_logical_cpuids = 1;
2243 
2244 /*
2245  * Used to store mapping between logical CPU IDs and APIC IDs.
2246  */
2247 static int cpuid_to_apicid[] = {
2248 	[0 ... NR_CPUS - 1] = -1,
2249 };
2250 
2251 #ifdef CONFIG_SMP
2252 /**
2253  * apic_id_is_primary_thread - Check whether APIC ID belongs to a primary thread
2254  * @id:	APIC ID to check
2255  */
2256 bool apic_id_is_primary_thread(unsigned int apicid)
2257 {
2258 	u32 mask;
2259 
2260 	if (smp_num_siblings == 1)
2261 		return true;
2262 	/* Isolate the SMT bit(s) in the APICID and check for 0 */
2263 	mask = (1U << (fls(smp_num_siblings) - 1)) - 1;
2264 	return !(apicid & mask);
2265 }
2266 #endif
2267 
2268 /*
2269  * Should use this API to allocate logical CPU IDs to keep nr_logical_cpuids
2270  * and cpuid_to_apicid[] synchronized.
2271  */
2272 static int allocate_logical_cpuid(int apicid)
2273 {
2274 	int i;
2275 
2276 	/*
2277 	 * cpuid <-> apicid mapping is persistent, so when a cpu is up,
2278 	 * check if the kernel has allocated a cpuid for it.
2279 	 */
2280 	for (i = 0; i < nr_logical_cpuids; i++) {
2281 		if (cpuid_to_apicid[i] == apicid)
2282 			return i;
2283 	}
2284 
2285 	/* Allocate a new cpuid. */
2286 	if (nr_logical_cpuids >= nr_cpu_ids) {
2287 		WARN_ONCE(1, "APIC: NR_CPUS/possible_cpus limit of %u reached. "
2288 			     "Processor %d/0x%x and the rest are ignored.\n",
2289 			     nr_cpu_ids, nr_logical_cpuids, apicid);
2290 		return -EINVAL;
2291 	}
2292 
2293 	cpuid_to_apicid[nr_logical_cpuids] = apicid;
2294 	return nr_logical_cpuids++;
2295 }
2296 
2297 int generic_processor_info(int apicid, int version)
2298 {
2299 	int cpu, max = nr_cpu_ids;
2300 	bool boot_cpu_detected = physid_isset(boot_cpu_physical_apicid,
2301 				phys_cpu_present_map);
2302 
2303 	/*
2304 	 * boot_cpu_physical_apicid is designed to have the apicid
2305 	 * returned by read_apic_id(), i.e, the apicid of the
2306 	 * currently booting-up processor. However, on some platforms,
2307 	 * it is temporarily modified by the apicid reported as BSP
2308 	 * through MP table. Concretely:
2309 	 *
2310 	 * - arch/x86/kernel/mpparse.c: MP_processor_info()
2311 	 * - arch/x86/mm/amdtopology.c: amd_numa_init()
2312 	 *
2313 	 * This function is executed with the modified
2314 	 * boot_cpu_physical_apicid. So, disabled_cpu_apicid kernel
2315 	 * parameter doesn't work to disable APs on kdump 2nd kernel.
2316 	 *
2317 	 * Since fixing handling of boot_cpu_physical_apicid requires
2318 	 * another discussion and tests on each platform, we leave it
2319 	 * for now and here we use read_apic_id() directly in this
2320 	 * function, generic_processor_info().
2321 	 */
2322 	if (disabled_cpu_apicid != BAD_APICID &&
2323 	    disabled_cpu_apicid != read_apic_id() &&
2324 	    disabled_cpu_apicid == apicid) {
2325 		int thiscpu = num_processors + disabled_cpus;
2326 
2327 		pr_warning("APIC: Disabling requested cpu."
2328 			   " Processor %d/0x%x ignored.\n",
2329 			   thiscpu, apicid);
2330 
2331 		disabled_cpus++;
2332 		return -ENODEV;
2333 	}
2334 
2335 	/*
2336 	 * If boot cpu has not been detected yet, then only allow upto
2337 	 * nr_cpu_ids - 1 processors and keep one slot free for boot cpu
2338 	 */
2339 	if (!boot_cpu_detected && num_processors >= nr_cpu_ids - 1 &&
2340 	    apicid != boot_cpu_physical_apicid) {
2341 		int thiscpu = max + disabled_cpus - 1;
2342 
2343 		pr_warning(
2344 			"APIC: NR_CPUS/possible_cpus limit of %i almost"
2345 			" reached. Keeping one slot for boot cpu."
2346 			"  Processor %d/0x%x ignored.\n", max, thiscpu, apicid);
2347 
2348 		disabled_cpus++;
2349 		return -ENODEV;
2350 	}
2351 
2352 	if (num_processors >= nr_cpu_ids) {
2353 		int thiscpu = max + disabled_cpus;
2354 
2355 		pr_warning("APIC: NR_CPUS/possible_cpus limit of %i "
2356 			   "reached. Processor %d/0x%x ignored.\n",
2357 			   max, thiscpu, apicid);
2358 
2359 		disabled_cpus++;
2360 		return -EINVAL;
2361 	}
2362 
2363 	if (apicid == boot_cpu_physical_apicid) {
2364 		/*
2365 		 * x86_bios_cpu_apicid is required to have processors listed
2366 		 * in same order as logical cpu numbers. Hence the first
2367 		 * entry is BSP, and so on.
2368 		 * boot_cpu_init() already hold bit 0 in cpu_present_mask
2369 		 * for BSP.
2370 		 */
2371 		cpu = 0;
2372 
2373 		/* Logical cpuid 0 is reserved for BSP. */
2374 		cpuid_to_apicid[0] = apicid;
2375 	} else {
2376 		cpu = allocate_logical_cpuid(apicid);
2377 		if (cpu < 0) {
2378 			disabled_cpus++;
2379 			return -EINVAL;
2380 		}
2381 	}
2382 
2383 	/*
2384 	 * Validate version
2385 	 */
2386 	if (version == 0x0) {
2387 		pr_warning("BIOS bug: APIC version is 0 for CPU %d/0x%x, fixing up to 0x10\n",
2388 			   cpu, apicid);
2389 		version = 0x10;
2390 	}
2391 
2392 	if (version != boot_cpu_apic_version) {
2393 		pr_warning("BIOS bug: APIC version mismatch, boot CPU: %x, CPU %d: version %x\n",
2394 			boot_cpu_apic_version, cpu, version);
2395 	}
2396 
2397 	if (apicid > max_physical_apicid)
2398 		max_physical_apicid = apicid;
2399 
2400 #if defined(CONFIG_SMP) || defined(CONFIG_X86_64)
2401 	early_per_cpu(x86_cpu_to_apicid, cpu) = apicid;
2402 	early_per_cpu(x86_bios_cpu_apicid, cpu) = apicid;
2403 #endif
2404 #ifdef CONFIG_X86_32
2405 	early_per_cpu(x86_cpu_to_logical_apicid, cpu) =
2406 		apic->x86_32_early_logical_apicid(cpu);
2407 #endif
2408 	set_cpu_possible(cpu, true);
2409 	physid_set(apicid, phys_cpu_present_map);
2410 	set_cpu_present(cpu, true);
2411 	num_processors++;
2412 
2413 	return cpu;
2414 }
2415 
2416 int hard_smp_processor_id(void)
2417 {
2418 	return read_apic_id();
2419 }
2420 
2421 /*
2422  * Override the generic EOI implementation with an optimized version.
2423  * Only called during early boot when only one CPU is active and with
2424  * interrupts disabled, so we know this does not race with actual APIC driver
2425  * use.
2426  */
2427 void __init apic_set_eoi_write(void (*eoi_write)(u32 reg, u32 v))
2428 {
2429 	struct apic **drv;
2430 
2431 	for (drv = __apicdrivers; drv < __apicdrivers_end; drv++) {
2432 		/* Should happen once for each apic */
2433 		WARN_ON((*drv)->eoi_write == eoi_write);
2434 		(*drv)->native_eoi_write = (*drv)->eoi_write;
2435 		(*drv)->eoi_write = eoi_write;
2436 	}
2437 }
2438 
2439 static void __init apic_bsp_up_setup(void)
2440 {
2441 #ifdef CONFIG_X86_64
2442 	apic_write(APIC_ID, apic->set_apic_id(boot_cpu_physical_apicid));
2443 #else
2444 	/*
2445 	 * Hack: In case of kdump, after a crash, kernel might be booting
2446 	 * on a cpu with non-zero lapic id. But boot_cpu_physical_apicid
2447 	 * might be zero if read from MP tables. Get it from LAPIC.
2448 	 */
2449 # ifdef CONFIG_CRASH_DUMP
2450 	boot_cpu_physical_apicid = read_apic_id();
2451 # endif
2452 #endif
2453 	physid_set_mask_of_physid(boot_cpu_physical_apicid, &phys_cpu_present_map);
2454 }
2455 
2456 /**
2457  * apic_bsp_setup - Setup function for local apic and io-apic
2458  * @upmode:		Force UP mode (for APIC_init_uniprocessor)
2459  */
2460 static void __init apic_bsp_setup(bool upmode)
2461 {
2462 	connect_bsp_APIC();
2463 	if (upmode)
2464 		apic_bsp_up_setup();
2465 	setup_local_APIC();
2466 
2467 	enable_IO_APIC();
2468 	end_local_APIC_setup();
2469 	irq_remap_enable_fault_handling();
2470 	setup_IO_APIC();
2471 }
2472 
2473 #ifdef CONFIG_UP_LATE_INIT
2474 void __init up_late_init(void)
2475 {
2476 	if (apic_intr_mode == APIC_PIC)
2477 		return;
2478 
2479 	/* Setup local timer */
2480 	x86_init.timers.setup_percpu_clockev();
2481 }
2482 #endif
2483 
2484 /*
2485  * Power management
2486  */
2487 #ifdef CONFIG_PM
2488 
2489 static struct {
2490 	/*
2491 	 * 'active' is true if the local APIC was enabled by us and
2492 	 * not the BIOS; this signifies that we are also responsible
2493 	 * for disabling it before entering apm/acpi suspend
2494 	 */
2495 	int active;
2496 	/* r/w apic fields */
2497 	unsigned int apic_id;
2498 	unsigned int apic_taskpri;
2499 	unsigned int apic_ldr;
2500 	unsigned int apic_dfr;
2501 	unsigned int apic_spiv;
2502 	unsigned int apic_lvtt;
2503 	unsigned int apic_lvtpc;
2504 	unsigned int apic_lvt0;
2505 	unsigned int apic_lvt1;
2506 	unsigned int apic_lvterr;
2507 	unsigned int apic_tmict;
2508 	unsigned int apic_tdcr;
2509 	unsigned int apic_thmr;
2510 	unsigned int apic_cmci;
2511 } apic_pm_state;
2512 
2513 static int lapic_suspend(void)
2514 {
2515 	unsigned long flags;
2516 	int maxlvt;
2517 
2518 	if (!apic_pm_state.active)
2519 		return 0;
2520 
2521 	maxlvt = lapic_get_maxlvt();
2522 
2523 	apic_pm_state.apic_id = apic_read(APIC_ID);
2524 	apic_pm_state.apic_taskpri = apic_read(APIC_TASKPRI);
2525 	apic_pm_state.apic_ldr = apic_read(APIC_LDR);
2526 	apic_pm_state.apic_dfr = apic_read(APIC_DFR);
2527 	apic_pm_state.apic_spiv = apic_read(APIC_SPIV);
2528 	apic_pm_state.apic_lvtt = apic_read(APIC_LVTT);
2529 	if (maxlvt >= 4)
2530 		apic_pm_state.apic_lvtpc = apic_read(APIC_LVTPC);
2531 	apic_pm_state.apic_lvt0 = apic_read(APIC_LVT0);
2532 	apic_pm_state.apic_lvt1 = apic_read(APIC_LVT1);
2533 	apic_pm_state.apic_lvterr = apic_read(APIC_LVTERR);
2534 	apic_pm_state.apic_tmict = apic_read(APIC_TMICT);
2535 	apic_pm_state.apic_tdcr = apic_read(APIC_TDCR);
2536 #ifdef CONFIG_X86_THERMAL_VECTOR
2537 	if (maxlvt >= 5)
2538 		apic_pm_state.apic_thmr = apic_read(APIC_LVTTHMR);
2539 #endif
2540 #ifdef CONFIG_X86_MCE_INTEL
2541 	if (maxlvt >= 6)
2542 		apic_pm_state.apic_cmci = apic_read(APIC_LVTCMCI);
2543 #endif
2544 
2545 	local_irq_save(flags);
2546 	disable_local_APIC();
2547 
2548 	irq_remapping_disable();
2549 
2550 	local_irq_restore(flags);
2551 	return 0;
2552 }
2553 
2554 static void lapic_resume(void)
2555 {
2556 	unsigned int l, h;
2557 	unsigned long flags;
2558 	int maxlvt;
2559 
2560 	if (!apic_pm_state.active)
2561 		return;
2562 
2563 	local_irq_save(flags);
2564 
2565 	/*
2566 	 * IO-APIC and PIC have their own resume routines.
2567 	 * We just mask them here to make sure the interrupt
2568 	 * subsystem is completely quiet while we enable x2apic
2569 	 * and interrupt-remapping.
2570 	 */
2571 	mask_ioapic_entries();
2572 	legacy_pic->mask_all();
2573 
2574 	if (x2apic_mode) {
2575 		__x2apic_enable();
2576 	} else {
2577 		/*
2578 		 * Make sure the APICBASE points to the right address
2579 		 *
2580 		 * FIXME! This will be wrong if we ever support suspend on
2581 		 * SMP! We'll need to do this as part of the CPU restore!
2582 		 */
2583 		if (boot_cpu_data.x86 >= 6) {
2584 			rdmsr(MSR_IA32_APICBASE, l, h);
2585 			l &= ~MSR_IA32_APICBASE_BASE;
2586 			l |= MSR_IA32_APICBASE_ENABLE | mp_lapic_addr;
2587 			wrmsr(MSR_IA32_APICBASE, l, h);
2588 		}
2589 	}
2590 
2591 	maxlvt = lapic_get_maxlvt();
2592 	apic_write(APIC_LVTERR, ERROR_APIC_VECTOR | APIC_LVT_MASKED);
2593 	apic_write(APIC_ID, apic_pm_state.apic_id);
2594 	apic_write(APIC_DFR, apic_pm_state.apic_dfr);
2595 	apic_write(APIC_LDR, apic_pm_state.apic_ldr);
2596 	apic_write(APIC_TASKPRI, apic_pm_state.apic_taskpri);
2597 	apic_write(APIC_SPIV, apic_pm_state.apic_spiv);
2598 	apic_write(APIC_LVT0, apic_pm_state.apic_lvt0);
2599 	apic_write(APIC_LVT1, apic_pm_state.apic_lvt1);
2600 #ifdef CONFIG_X86_THERMAL_VECTOR
2601 	if (maxlvt >= 5)
2602 		apic_write(APIC_LVTTHMR, apic_pm_state.apic_thmr);
2603 #endif
2604 #ifdef CONFIG_X86_MCE_INTEL
2605 	if (maxlvt >= 6)
2606 		apic_write(APIC_LVTCMCI, apic_pm_state.apic_cmci);
2607 #endif
2608 	if (maxlvt >= 4)
2609 		apic_write(APIC_LVTPC, apic_pm_state.apic_lvtpc);
2610 	apic_write(APIC_LVTT, apic_pm_state.apic_lvtt);
2611 	apic_write(APIC_TDCR, apic_pm_state.apic_tdcr);
2612 	apic_write(APIC_TMICT, apic_pm_state.apic_tmict);
2613 	apic_write(APIC_ESR, 0);
2614 	apic_read(APIC_ESR);
2615 	apic_write(APIC_LVTERR, apic_pm_state.apic_lvterr);
2616 	apic_write(APIC_ESR, 0);
2617 	apic_read(APIC_ESR);
2618 
2619 	irq_remapping_reenable(x2apic_mode);
2620 
2621 	local_irq_restore(flags);
2622 }
2623 
2624 /*
2625  * This device has no shutdown method - fully functioning local APICs
2626  * are needed on every CPU up until machine_halt/restart/poweroff.
2627  */
2628 
2629 static struct syscore_ops lapic_syscore_ops = {
2630 	.resume		= lapic_resume,
2631 	.suspend	= lapic_suspend,
2632 };
2633 
2634 static void apic_pm_activate(void)
2635 {
2636 	apic_pm_state.active = 1;
2637 }
2638 
2639 static int __init init_lapic_sysfs(void)
2640 {
2641 	/* XXX: remove suspend/resume procs if !apic_pm_state.active? */
2642 	if (boot_cpu_has(X86_FEATURE_APIC))
2643 		register_syscore_ops(&lapic_syscore_ops);
2644 
2645 	return 0;
2646 }
2647 
2648 /* local apic needs to resume before other devices access its registers. */
2649 core_initcall(init_lapic_sysfs);
2650 
2651 #else	/* CONFIG_PM */
2652 
2653 static void apic_pm_activate(void) { }
2654 
2655 #endif	/* CONFIG_PM */
2656 
2657 #ifdef CONFIG_X86_64
2658 
2659 static int multi_checked;
2660 static int multi;
2661 
2662 static int set_multi(const struct dmi_system_id *d)
2663 {
2664 	if (multi)
2665 		return 0;
2666 	pr_info("APIC: %s detected, Multi Chassis\n", d->ident);
2667 	multi = 1;
2668 	return 0;
2669 }
2670 
2671 static const struct dmi_system_id multi_dmi_table[] = {
2672 	{
2673 		.callback = set_multi,
2674 		.ident = "IBM System Summit2",
2675 		.matches = {
2676 			DMI_MATCH(DMI_SYS_VENDOR, "IBM"),
2677 			DMI_MATCH(DMI_PRODUCT_NAME, "Summit2"),
2678 		},
2679 	},
2680 	{}
2681 };
2682 
2683 static void dmi_check_multi(void)
2684 {
2685 	if (multi_checked)
2686 		return;
2687 
2688 	dmi_check_system(multi_dmi_table);
2689 	multi_checked = 1;
2690 }
2691 
2692 /*
2693  * apic_is_clustered_box() -- Check if we can expect good TSC
2694  *
2695  * Thus far, the major user of this is IBM's Summit2 series:
2696  * Clustered boxes may have unsynced TSC problems if they are
2697  * multi-chassis.
2698  * Use DMI to check them
2699  */
2700 int apic_is_clustered_box(void)
2701 {
2702 	dmi_check_multi();
2703 	return multi;
2704 }
2705 #endif
2706 
2707 /*
2708  * APIC command line parameters
2709  */
2710 static int __init setup_disableapic(char *arg)
2711 {
2712 	disable_apic = 1;
2713 	setup_clear_cpu_cap(X86_FEATURE_APIC);
2714 	return 0;
2715 }
2716 early_param("disableapic", setup_disableapic);
2717 
2718 /* same as disableapic, for compatibility */
2719 static int __init setup_nolapic(char *arg)
2720 {
2721 	return setup_disableapic(arg);
2722 }
2723 early_param("nolapic", setup_nolapic);
2724 
2725 static int __init parse_lapic_timer_c2_ok(char *arg)
2726 {
2727 	local_apic_timer_c2_ok = 1;
2728 	return 0;
2729 }
2730 early_param("lapic_timer_c2_ok", parse_lapic_timer_c2_ok);
2731 
2732 static int __init parse_disable_apic_timer(char *arg)
2733 {
2734 	disable_apic_timer = 1;
2735 	return 0;
2736 }
2737 early_param("noapictimer", parse_disable_apic_timer);
2738 
2739 static int __init parse_nolapic_timer(char *arg)
2740 {
2741 	disable_apic_timer = 1;
2742 	return 0;
2743 }
2744 early_param("nolapic_timer", parse_nolapic_timer);
2745 
2746 static int __init apic_set_verbosity(char *arg)
2747 {
2748 	if (!arg)  {
2749 #ifdef CONFIG_X86_64
2750 		skip_ioapic_setup = 0;
2751 		return 0;
2752 #endif
2753 		return -EINVAL;
2754 	}
2755 
2756 	if (strcmp("debug", arg) == 0)
2757 		apic_verbosity = APIC_DEBUG;
2758 	else if (strcmp("verbose", arg) == 0)
2759 		apic_verbosity = APIC_VERBOSE;
2760 #ifdef CONFIG_X86_64
2761 	else {
2762 		pr_warning("APIC Verbosity level %s not recognised"
2763 			" use apic=verbose or apic=debug\n", arg);
2764 		return -EINVAL;
2765 	}
2766 #endif
2767 
2768 	return 0;
2769 }
2770 early_param("apic", apic_set_verbosity);
2771 
2772 static int __init lapic_insert_resource(void)
2773 {
2774 	if (!apic_phys)
2775 		return -1;
2776 
2777 	/* Put local APIC into the resource map. */
2778 	lapic_resource.start = apic_phys;
2779 	lapic_resource.end = lapic_resource.start + PAGE_SIZE - 1;
2780 	insert_resource(&iomem_resource, &lapic_resource);
2781 
2782 	return 0;
2783 }
2784 
2785 /*
2786  * need call insert after e820__reserve_resources()
2787  * that is using request_resource
2788  */
2789 late_initcall(lapic_insert_resource);
2790 
2791 static int __init apic_set_disabled_cpu_apicid(char *arg)
2792 {
2793 	if (!arg || !get_option(&arg, &disabled_cpu_apicid))
2794 		return -EINVAL;
2795 
2796 	return 0;
2797 }
2798 early_param("disable_cpu_apicid", apic_set_disabled_cpu_apicid);
2799 
2800 static int __init apic_set_extnmi(char *arg)
2801 {
2802 	if (!arg)
2803 		return -EINVAL;
2804 
2805 	if (!strncmp("all", arg, 3))
2806 		apic_extnmi = APIC_EXTNMI_ALL;
2807 	else if (!strncmp("none", arg, 4))
2808 		apic_extnmi = APIC_EXTNMI_NONE;
2809 	else if (!strncmp("bsp", arg, 3))
2810 		apic_extnmi = APIC_EXTNMI_BSP;
2811 	else {
2812 		pr_warn("Unknown external NMI delivery mode `%s' ignored\n", arg);
2813 		return -EINVAL;
2814 	}
2815 
2816 	return 0;
2817 }
2818 early_param("apic_extnmi", apic_set_extnmi);
2819