xref: /linux/arch/x86/kvm/lapic.c (revision f2ee442115c9b6219083c019939a9cc0c9abb2f8)
1 
2 /*
3  * Local APIC virtualization
4  *
5  * Copyright (C) 2006 Qumranet, Inc.
6  * Copyright (C) 2007 Novell
7  * Copyright (C) 2007 Intel
8  * Copyright 2009 Red Hat, Inc. and/or its affiliates.
9  *
10  * Authors:
11  *   Dor Laor <dor.laor@qumranet.com>
12  *   Gregory Haskins <ghaskins@novell.com>
13  *   Yaozu (Eddie) Dong <eddie.dong@intel.com>
14  *
15  * Based on Xen 3.1 code, Copyright (c) 2004, Intel Corporation.
16  *
17  * This work is licensed under the terms of the GNU GPL, version 2.  See
18  * the COPYING file in the top-level directory.
19  */
20 
21 #include <linux/kvm_host.h>
22 #include <linux/kvm.h>
23 #include <linux/mm.h>
24 #include <linux/highmem.h>
25 #include <linux/smp.h>
26 #include <linux/hrtimer.h>
27 #include <linux/io.h>
28 #include <linux/module.h>
29 #include <linux/math64.h>
30 #include <linux/slab.h>
31 #include <asm/processor.h>
32 #include <asm/msr.h>
33 #include <asm/page.h>
34 #include <asm/current.h>
35 #include <asm/apicdef.h>
36 #include <linux/atomic.h>
37 #include "kvm_cache_regs.h"
38 #include "irq.h"
39 #include "trace.h"
40 #include "x86.h"
41 
42 #ifndef CONFIG_X86_64
43 #define mod_64(x, y) ((x) - (y) * div64_u64(x, y))
44 #else
45 #define mod_64(x, y) ((x) % (y))
46 #endif
47 
48 #define PRId64 "d"
49 #define PRIx64 "llx"
50 #define PRIu64 "u"
51 #define PRIo64 "o"
52 
53 #define APIC_BUS_CYCLE_NS 1
54 
55 /* #define apic_debug(fmt,arg...) printk(KERN_WARNING fmt,##arg) */
56 #define apic_debug(fmt, arg...)
57 
58 #define APIC_LVT_NUM			6
59 /* 14 is the version for Xeon and Pentium 8.4.8*/
60 #define APIC_VERSION			(0x14UL | ((APIC_LVT_NUM - 1) << 16))
61 #define LAPIC_MMIO_LENGTH		(1 << 12)
62 /* followed define is not in apicdef.h */
63 #define APIC_SHORT_MASK			0xc0000
64 #define APIC_DEST_NOSHORT		0x0
65 #define APIC_DEST_MASK			0x800
66 #define MAX_APIC_VECTOR			256
67 
68 #define VEC_POS(v) ((v) & (32 - 1))
69 #define REG_POS(v) (((v) >> 5) << 4)
70 
71 static unsigned int min_timer_period_us = 500;
72 module_param(min_timer_period_us, uint, S_IRUGO | S_IWUSR);
73 
74 static inline u32 apic_get_reg(struct kvm_lapic *apic, int reg_off)
75 {
76 	return *((u32 *) (apic->regs + reg_off));
77 }
78 
79 static inline void apic_set_reg(struct kvm_lapic *apic, int reg_off, u32 val)
80 {
81 	*((u32 *) (apic->regs + reg_off)) = val;
82 }
83 
84 static inline int apic_test_and_set_vector(int vec, void *bitmap)
85 {
86 	return test_and_set_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
87 }
88 
89 static inline int apic_test_and_clear_vector(int vec, void *bitmap)
90 {
91 	return test_and_clear_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
92 }
93 
94 static inline void apic_set_vector(int vec, void *bitmap)
95 {
96 	set_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
97 }
98 
99 static inline void apic_clear_vector(int vec, void *bitmap)
100 {
101 	clear_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
102 }
103 
104 static inline int apic_hw_enabled(struct kvm_lapic *apic)
105 {
106 	return (apic)->vcpu->arch.apic_base & MSR_IA32_APICBASE_ENABLE;
107 }
108 
109 static inline int  apic_sw_enabled(struct kvm_lapic *apic)
110 {
111 	return apic_get_reg(apic, APIC_SPIV) & APIC_SPIV_APIC_ENABLED;
112 }
113 
114 static inline int apic_enabled(struct kvm_lapic *apic)
115 {
116 	return apic_sw_enabled(apic) &&	apic_hw_enabled(apic);
117 }
118 
119 #define LVT_MASK	\
120 	(APIC_LVT_MASKED | APIC_SEND_PENDING | APIC_VECTOR_MASK)
121 
122 #define LINT_MASK	\
123 	(LVT_MASK | APIC_MODE_MASK | APIC_INPUT_POLARITY | \
124 	 APIC_LVT_REMOTE_IRR | APIC_LVT_LEVEL_TRIGGER)
125 
126 static inline int kvm_apic_id(struct kvm_lapic *apic)
127 {
128 	return (apic_get_reg(apic, APIC_ID) >> 24) & 0xff;
129 }
130 
131 static inline int apic_lvt_enabled(struct kvm_lapic *apic, int lvt_type)
132 {
133 	return !(apic_get_reg(apic, lvt_type) & APIC_LVT_MASKED);
134 }
135 
136 static inline int apic_lvt_vector(struct kvm_lapic *apic, int lvt_type)
137 {
138 	return apic_get_reg(apic, lvt_type) & APIC_VECTOR_MASK;
139 }
140 
141 static inline int apic_lvtt_oneshot(struct kvm_lapic *apic)
142 {
143 	return ((apic_get_reg(apic, APIC_LVTT) &
144 		apic->lapic_timer.timer_mode_mask) == APIC_LVT_TIMER_ONESHOT);
145 }
146 
147 static inline int apic_lvtt_period(struct kvm_lapic *apic)
148 {
149 	return ((apic_get_reg(apic, APIC_LVTT) &
150 		apic->lapic_timer.timer_mode_mask) == APIC_LVT_TIMER_PERIODIC);
151 }
152 
153 static inline int apic_lvtt_tscdeadline(struct kvm_lapic *apic)
154 {
155 	return ((apic_get_reg(apic, APIC_LVTT) &
156 		apic->lapic_timer.timer_mode_mask) ==
157 			APIC_LVT_TIMER_TSCDEADLINE);
158 }
159 
160 static inline int apic_lvt_nmi_mode(u32 lvt_val)
161 {
162 	return (lvt_val & (APIC_MODE_MASK | APIC_LVT_MASKED)) == APIC_DM_NMI;
163 }
164 
165 void kvm_apic_set_version(struct kvm_vcpu *vcpu)
166 {
167 	struct kvm_lapic *apic = vcpu->arch.apic;
168 	struct kvm_cpuid_entry2 *feat;
169 	u32 v = APIC_VERSION;
170 
171 	if (!irqchip_in_kernel(vcpu->kvm))
172 		return;
173 
174 	feat = kvm_find_cpuid_entry(apic->vcpu, 0x1, 0);
175 	if (feat && (feat->ecx & (1 << (X86_FEATURE_X2APIC & 31))))
176 		v |= APIC_LVR_DIRECTED_EOI;
177 	apic_set_reg(apic, APIC_LVR, v);
178 }
179 
180 static inline int apic_x2apic_mode(struct kvm_lapic *apic)
181 {
182 	return apic->vcpu->arch.apic_base & X2APIC_ENABLE;
183 }
184 
185 static unsigned int apic_lvt_mask[APIC_LVT_NUM] = {
186 	LVT_MASK ,      /* part LVTT mask, timer mode mask added at runtime */
187 	LVT_MASK | APIC_MODE_MASK,	/* LVTTHMR */
188 	LVT_MASK | APIC_MODE_MASK,	/* LVTPC */
189 	LINT_MASK, LINT_MASK,	/* LVT0-1 */
190 	LVT_MASK		/* LVTERR */
191 };
192 
193 static int find_highest_vector(void *bitmap)
194 {
195 	u32 *word = bitmap;
196 	int word_offset = MAX_APIC_VECTOR >> 5;
197 
198 	while ((word_offset != 0) && (word[(--word_offset) << 2] == 0))
199 		continue;
200 
201 	if (likely(!word_offset && !word[0]))
202 		return -1;
203 	else
204 		return fls(word[word_offset << 2]) - 1 + (word_offset << 5);
205 }
206 
207 static inline int apic_test_and_set_irr(int vec, struct kvm_lapic *apic)
208 {
209 	apic->irr_pending = true;
210 	return apic_test_and_set_vector(vec, apic->regs + APIC_IRR);
211 }
212 
213 static inline int apic_search_irr(struct kvm_lapic *apic)
214 {
215 	return find_highest_vector(apic->regs + APIC_IRR);
216 }
217 
218 static inline int apic_find_highest_irr(struct kvm_lapic *apic)
219 {
220 	int result;
221 
222 	if (!apic->irr_pending)
223 		return -1;
224 
225 	result = apic_search_irr(apic);
226 	ASSERT(result == -1 || result >= 16);
227 
228 	return result;
229 }
230 
231 static inline void apic_clear_irr(int vec, struct kvm_lapic *apic)
232 {
233 	apic->irr_pending = false;
234 	apic_clear_vector(vec, apic->regs + APIC_IRR);
235 	if (apic_search_irr(apic) != -1)
236 		apic->irr_pending = true;
237 }
238 
239 int kvm_lapic_find_highest_irr(struct kvm_vcpu *vcpu)
240 {
241 	struct kvm_lapic *apic = vcpu->arch.apic;
242 	int highest_irr;
243 
244 	/* This may race with setting of irr in __apic_accept_irq() and
245 	 * value returned may be wrong, but kvm_vcpu_kick() in __apic_accept_irq
246 	 * will cause vmexit immediately and the value will be recalculated
247 	 * on the next vmentry.
248 	 */
249 	if (!apic)
250 		return 0;
251 	highest_irr = apic_find_highest_irr(apic);
252 
253 	return highest_irr;
254 }
255 
256 static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
257 			     int vector, int level, int trig_mode);
258 
259 int kvm_apic_set_irq(struct kvm_vcpu *vcpu, struct kvm_lapic_irq *irq)
260 {
261 	struct kvm_lapic *apic = vcpu->arch.apic;
262 
263 	return __apic_accept_irq(apic, irq->delivery_mode, irq->vector,
264 			irq->level, irq->trig_mode);
265 }
266 
267 static inline int apic_find_highest_isr(struct kvm_lapic *apic)
268 {
269 	int result;
270 
271 	result = find_highest_vector(apic->regs + APIC_ISR);
272 	ASSERT(result == -1 || result >= 16);
273 
274 	return result;
275 }
276 
277 static void apic_update_ppr(struct kvm_lapic *apic)
278 {
279 	u32 tpr, isrv, ppr, old_ppr;
280 	int isr;
281 
282 	old_ppr = apic_get_reg(apic, APIC_PROCPRI);
283 	tpr = apic_get_reg(apic, APIC_TASKPRI);
284 	isr = apic_find_highest_isr(apic);
285 	isrv = (isr != -1) ? isr : 0;
286 
287 	if ((tpr & 0xf0) >= (isrv & 0xf0))
288 		ppr = tpr & 0xff;
289 	else
290 		ppr = isrv & 0xf0;
291 
292 	apic_debug("vlapic %p, ppr 0x%x, isr 0x%x, isrv 0x%x",
293 		   apic, ppr, isr, isrv);
294 
295 	if (old_ppr != ppr) {
296 		apic_set_reg(apic, APIC_PROCPRI, ppr);
297 		if (ppr < old_ppr)
298 			kvm_make_request(KVM_REQ_EVENT, apic->vcpu);
299 	}
300 }
301 
302 static void apic_set_tpr(struct kvm_lapic *apic, u32 tpr)
303 {
304 	apic_set_reg(apic, APIC_TASKPRI, tpr);
305 	apic_update_ppr(apic);
306 }
307 
308 int kvm_apic_match_physical_addr(struct kvm_lapic *apic, u16 dest)
309 {
310 	return dest == 0xff || kvm_apic_id(apic) == dest;
311 }
312 
313 int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
314 {
315 	int result = 0;
316 	u32 logical_id;
317 
318 	if (apic_x2apic_mode(apic)) {
319 		logical_id = apic_get_reg(apic, APIC_LDR);
320 		return logical_id & mda;
321 	}
322 
323 	logical_id = GET_APIC_LOGICAL_ID(apic_get_reg(apic, APIC_LDR));
324 
325 	switch (apic_get_reg(apic, APIC_DFR)) {
326 	case APIC_DFR_FLAT:
327 		if (logical_id & mda)
328 			result = 1;
329 		break;
330 	case APIC_DFR_CLUSTER:
331 		if (((logical_id >> 4) == (mda >> 0x4))
332 		    && (logical_id & mda & 0xf))
333 			result = 1;
334 		break;
335 	default:
336 		apic_debug("Bad DFR vcpu %d: %08x\n",
337 			   apic->vcpu->vcpu_id, apic_get_reg(apic, APIC_DFR));
338 		break;
339 	}
340 
341 	return result;
342 }
343 
344 int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
345 			   int short_hand, int dest, int dest_mode)
346 {
347 	int result = 0;
348 	struct kvm_lapic *target = vcpu->arch.apic;
349 
350 	apic_debug("target %p, source %p, dest 0x%x, "
351 		   "dest_mode 0x%x, short_hand 0x%x\n",
352 		   target, source, dest, dest_mode, short_hand);
353 
354 	ASSERT(target);
355 	switch (short_hand) {
356 	case APIC_DEST_NOSHORT:
357 		if (dest_mode == 0)
358 			/* Physical mode. */
359 			result = kvm_apic_match_physical_addr(target, dest);
360 		else
361 			/* Logical mode. */
362 			result = kvm_apic_match_logical_addr(target, dest);
363 		break;
364 	case APIC_DEST_SELF:
365 		result = (target == source);
366 		break;
367 	case APIC_DEST_ALLINC:
368 		result = 1;
369 		break;
370 	case APIC_DEST_ALLBUT:
371 		result = (target != source);
372 		break;
373 	default:
374 		apic_debug("kvm: apic: Bad dest shorthand value %x\n",
375 			   short_hand);
376 		break;
377 	}
378 
379 	return result;
380 }
381 
382 /*
383  * Add a pending IRQ into lapic.
384  * Return 1 if successfully added and 0 if discarded.
385  */
386 static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
387 			     int vector, int level, int trig_mode)
388 {
389 	int result = 0;
390 	struct kvm_vcpu *vcpu = apic->vcpu;
391 
392 	switch (delivery_mode) {
393 	case APIC_DM_LOWEST:
394 		vcpu->arch.apic_arb_prio++;
395 	case APIC_DM_FIXED:
396 		/* FIXME add logic for vcpu on reset */
397 		if (unlikely(!apic_enabled(apic)))
398 			break;
399 
400 		if (trig_mode) {
401 			apic_debug("level trig mode for vector %d", vector);
402 			apic_set_vector(vector, apic->regs + APIC_TMR);
403 		} else
404 			apic_clear_vector(vector, apic->regs + APIC_TMR);
405 
406 		result = !apic_test_and_set_irr(vector, apic);
407 		trace_kvm_apic_accept_irq(vcpu->vcpu_id, delivery_mode,
408 					  trig_mode, vector, !result);
409 		if (!result) {
410 			if (trig_mode)
411 				apic_debug("level trig mode repeatedly for "
412 						"vector %d", vector);
413 			break;
414 		}
415 
416 		kvm_make_request(KVM_REQ_EVENT, vcpu);
417 		kvm_vcpu_kick(vcpu);
418 		break;
419 
420 	case APIC_DM_REMRD:
421 		apic_debug("Ignoring delivery mode 3\n");
422 		break;
423 
424 	case APIC_DM_SMI:
425 		apic_debug("Ignoring guest SMI\n");
426 		break;
427 
428 	case APIC_DM_NMI:
429 		result = 1;
430 		kvm_inject_nmi(vcpu);
431 		kvm_vcpu_kick(vcpu);
432 		break;
433 
434 	case APIC_DM_INIT:
435 		if (level) {
436 			result = 1;
437 			vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
438 			kvm_make_request(KVM_REQ_EVENT, vcpu);
439 			kvm_vcpu_kick(vcpu);
440 		} else {
441 			apic_debug("Ignoring de-assert INIT to vcpu %d\n",
442 				   vcpu->vcpu_id);
443 		}
444 		break;
445 
446 	case APIC_DM_STARTUP:
447 		apic_debug("SIPI to vcpu %d vector 0x%02x\n",
448 			   vcpu->vcpu_id, vector);
449 		if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) {
450 			result = 1;
451 			vcpu->arch.sipi_vector = vector;
452 			vcpu->arch.mp_state = KVM_MP_STATE_SIPI_RECEIVED;
453 			kvm_make_request(KVM_REQ_EVENT, vcpu);
454 			kvm_vcpu_kick(vcpu);
455 		}
456 		break;
457 
458 	case APIC_DM_EXTINT:
459 		/*
460 		 * Should only be called by kvm_apic_local_deliver() with LVT0,
461 		 * before NMI watchdog was enabled. Already handled by
462 		 * kvm_apic_accept_pic_intr().
463 		 */
464 		break;
465 
466 	default:
467 		printk(KERN_ERR "TODO: unsupported delivery mode %x\n",
468 		       delivery_mode);
469 		break;
470 	}
471 	return result;
472 }
473 
474 int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2)
475 {
476 	return vcpu1->arch.apic_arb_prio - vcpu2->arch.apic_arb_prio;
477 }
478 
479 static void apic_set_eoi(struct kvm_lapic *apic)
480 {
481 	int vector = apic_find_highest_isr(apic);
482 	int trigger_mode;
483 	/*
484 	 * Not every write EOI will has corresponding ISR,
485 	 * one example is when Kernel check timer on setup_IO_APIC
486 	 */
487 	if (vector == -1)
488 		return;
489 
490 	apic_clear_vector(vector, apic->regs + APIC_ISR);
491 	apic_update_ppr(apic);
492 
493 	if (apic_test_and_clear_vector(vector, apic->regs + APIC_TMR))
494 		trigger_mode = IOAPIC_LEVEL_TRIG;
495 	else
496 		trigger_mode = IOAPIC_EDGE_TRIG;
497 	if (!(apic_get_reg(apic, APIC_SPIV) & APIC_SPIV_DIRECTED_EOI))
498 		kvm_ioapic_update_eoi(apic->vcpu->kvm, vector, trigger_mode);
499 	kvm_make_request(KVM_REQ_EVENT, apic->vcpu);
500 }
501 
502 static void apic_send_ipi(struct kvm_lapic *apic)
503 {
504 	u32 icr_low = apic_get_reg(apic, APIC_ICR);
505 	u32 icr_high = apic_get_reg(apic, APIC_ICR2);
506 	struct kvm_lapic_irq irq;
507 
508 	irq.vector = icr_low & APIC_VECTOR_MASK;
509 	irq.delivery_mode = icr_low & APIC_MODE_MASK;
510 	irq.dest_mode = icr_low & APIC_DEST_MASK;
511 	irq.level = icr_low & APIC_INT_ASSERT;
512 	irq.trig_mode = icr_low & APIC_INT_LEVELTRIG;
513 	irq.shorthand = icr_low & APIC_SHORT_MASK;
514 	if (apic_x2apic_mode(apic))
515 		irq.dest_id = icr_high;
516 	else
517 		irq.dest_id = GET_APIC_DEST_FIELD(icr_high);
518 
519 	trace_kvm_apic_ipi(icr_low, irq.dest_id);
520 
521 	apic_debug("icr_high 0x%x, icr_low 0x%x, "
522 		   "short_hand 0x%x, dest 0x%x, trig_mode 0x%x, level 0x%x, "
523 		   "dest_mode 0x%x, delivery_mode 0x%x, vector 0x%x\n",
524 		   icr_high, icr_low, irq.shorthand, irq.dest_id,
525 		   irq.trig_mode, irq.level, irq.dest_mode, irq.delivery_mode,
526 		   irq.vector);
527 
528 	kvm_irq_delivery_to_apic(apic->vcpu->kvm, apic, &irq);
529 }
530 
531 static u32 apic_get_tmcct(struct kvm_lapic *apic)
532 {
533 	ktime_t remaining;
534 	s64 ns;
535 	u32 tmcct;
536 
537 	ASSERT(apic != NULL);
538 
539 	/* if initial count is 0, current count should also be 0 */
540 	if (apic_get_reg(apic, APIC_TMICT) == 0)
541 		return 0;
542 
543 	remaining = hrtimer_get_remaining(&apic->lapic_timer.timer);
544 	if (ktime_to_ns(remaining) < 0)
545 		remaining = ktime_set(0, 0);
546 
547 	ns = mod_64(ktime_to_ns(remaining), apic->lapic_timer.period);
548 	tmcct = div64_u64(ns,
549 			 (APIC_BUS_CYCLE_NS * apic->divide_count));
550 
551 	return tmcct;
552 }
553 
554 static void __report_tpr_access(struct kvm_lapic *apic, bool write)
555 {
556 	struct kvm_vcpu *vcpu = apic->vcpu;
557 	struct kvm_run *run = vcpu->run;
558 
559 	kvm_make_request(KVM_REQ_REPORT_TPR_ACCESS, vcpu);
560 	run->tpr_access.rip = kvm_rip_read(vcpu);
561 	run->tpr_access.is_write = write;
562 }
563 
564 static inline void report_tpr_access(struct kvm_lapic *apic, bool write)
565 {
566 	if (apic->vcpu->arch.tpr_access_reporting)
567 		__report_tpr_access(apic, write);
568 }
569 
570 static u32 __apic_read(struct kvm_lapic *apic, unsigned int offset)
571 {
572 	u32 val = 0;
573 
574 	if (offset >= LAPIC_MMIO_LENGTH)
575 		return 0;
576 
577 	switch (offset) {
578 	case APIC_ID:
579 		if (apic_x2apic_mode(apic))
580 			val = kvm_apic_id(apic);
581 		else
582 			val = kvm_apic_id(apic) << 24;
583 		break;
584 	case APIC_ARBPRI:
585 		apic_debug("Access APIC ARBPRI register which is for P6\n");
586 		break;
587 
588 	case APIC_TMCCT:	/* Timer CCR */
589 		if (apic_lvtt_tscdeadline(apic))
590 			return 0;
591 
592 		val = apic_get_tmcct(apic);
593 		break;
594 
595 	case APIC_TASKPRI:
596 		report_tpr_access(apic, false);
597 		/* fall thru */
598 	default:
599 		apic_update_ppr(apic);
600 		val = apic_get_reg(apic, offset);
601 		break;
602 	}
603 
604 	return val;
605 }
606 
607 static inline struct kvm_lapic *to_lapic(struct kvm_io_device *dev)
608 {
609 	return container_of(dev, struct kvm_lapic, dev);
610 }
611 
612 static int apic_reg_read(struct kvm_lapic *apic, u32 offset, int len,
613 		void *data)
614 {
615 	unsigned char alignment = offset & 0xf;
616 	u32 result;
617 	/* this bitmask has a bit cleared for each reserver register */
618 	static const u64 rmask = 0x43ff01ffffffe70cULL;
619 
620 	if ((alignment + len) > 4) {
621 		apic_debug("KVM_APIC_READ: alignment error %x %d\n",
622 			   offset, len);
623 		return 1;
624 	}
625 
626 	if (offset > 0x3f0 || !(rmask & (1ULL << (offset >> 4)))) {
627 		apic_debug("KVM_APIC_READ: read reserved register %x\n",
628 			   offset);
629 		return 1;
630 	}
631 
632 	result = __apic_read(apic, offset & ~0xf);
633 
634 	trace_kvm_apic_read(offset, result);
635 
636 	switch (len) {
637 	case 1:
638 	case 2:
639 	case 4:
640 		memcpy(data, (char *)&result + alignment, len);
641 		break;
642 	default:
643 		printk(KERN_ERR "Local APIC read with len = %x, "
644 		       "should be 1,2, or 4 instead\n", len);
645 		break;
646 	}
647 	return 0;
648 }
649 
650 static int apic_mmio_in_range(struct kvm_lapic *apic, gpa_t addr)
651 {
652 	return apic_hw_enabled(apic) &&
653 	    addr >= apic->base_address &&
654 	    addr < apic->base_address + LAPIC_MMIO_LENGTH;
655 }
656 
657 static int apic_mmio_read(struct kvm_io_device *this,
658 			   gpa_t address, int len, void *data)
659 {
660 	struct kvm_lapic *apic = to_lapic(this);
661 	u32 offset = address - apic->base_address;
662 
663 	if (!apic_mmio_in_range(apic, address))
664 		return -EOPNOTSUPP;
665 
666 	apic_reg_read(apic, offset, len, data);
667 
668 	return 0;
669 }
670 
671 static void update_divide_count(struct kvm_lapic *apic)
672 {
673 	u32 tmp1, tmp2, tdcr;
674 
675 	tdcr = apic_get_reg(apic, APIC_TDCR);
676 	tmp1 = tdcr & 0xf;
677 	tmp2 = ((tmp1 & 0x3) | ((tmp1 & 0x8) >> 1)) + 1;
678 	apic->divide_count = 0x1 << (tmp2 & 0x7);
679 
680 	apic_debug("timer divide count is 0x%x\n",
681 				   apic->divide_count);
682 }
683 
684 static void start_apic_timer(struct kvm_lapic *apic)
685 {
686 	ktime_t now;
687 	atomic_set(&apic->lapic_timer.pending, 0);
688 
689 	if (apic_lvtt_period(apic) || apic_lvtt_oneshot(apic)) {
690 		/* lapic timer in oneshot or peroidic mode */
691 		now = apic->lapic_timer.timer.base->get_time();
692 		apic->lapic_timer.period = (u64)apic_get_reg(apic, APIC_TMICT)
693 			    * APIC_BUS_CYCLE_NS * apic->divide_count;
694 
695 		if (!apic->lapic_timer.period)
696 			return;
697 		/*
698 		 * Do not allow the guest to program periodic timers with small
699 		 * interval, since the hrtimers are not throttled by the host
700 		 * scheduler.
701 		 */
702 		if (apic_lvtt_period(apic)) {
703 			s64 min_period = min_timer_period_us * 1000LL;
704 
705 			if (apic->lapic_timer.period < min_period) {
706 				pr_info_ratelimited(
707 				    "kvm: vcpu %i: requested %lld ns "
708 				    "lapic timer period limited to %lld ns\n",
709 				    apic->vcpu->vcpu_id,
710 				    apic->lapic_timer.period, min_period);
711 				apic->lapic_timer.period = min_period;
712 			}
713 		}
714 
715 		hrtimer_start(&apic->lapic_timer.timer,
716 			      ktime_add_ns(now, apic->lapic_timer.period),
717 			      HRTIMER_MODE_ABS);
718 
719 		apic_debug("%s: bus cycle is %" PRId64 "ns, now 0x%016"
720 			   PRIx64 ", "
721 			   "timer initial count 0x%x, period %lldns, "
722 			   "expire @ 0x%016" PRIx64 ".\n", __func__,
723 			   APIC_BUS_CYCLE_NS, ktime_to_ns(now),
724 			   apic_get_reg(apic, APIC_TMICT),
725 			   apic->lapic_timer.period,
726 			   ktime_to_ns(ktime_add_ns(now,
727 					apic->lapic_timer.period)));
728 	} else if (apic_lvtt_tscdeadline(apic)) {
729 		/* lapic timer in tsc deadline mode */
730 		u64 guest_tsc, tscdeadline = apic->lapic_timer.tscdeadline;
731 		u64 ns = 0;
732 		struct kvm_vcpu *vcpu = apic->vcpu;
733 		unsigned long this_tsc_khz = vcpu_tsc_khz(vcpu);
734 		unsigned long flags;
735 
736 		if (unlikely(!tscdeadline || !this_tsc_khz))
737 			return;
738 
739 		local_irq_save(flags);
740 
741 		now = apic->lapic_timer.timer.base->get_time();
742 		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu);
743 		if (likely(tscdeadline > guest_tsc)) {
744 			ns = (tscdeadline - guest_tsc) * 1000000ULL;
745 			do_div(ns, this_tsc_khz);
746 		}
747 		hrtimer_start(&apic->lapic_timer.timer,
748 			ktime_add_ns(now, ns), HRTIMER_MODE_ABS);
749 
750 		local_irq_restore(flags);
751 	}
752 }
753 
754 static void apic_manage_nmi_watchdog(struct kvm_lapic *apic, u32 lvt0_val)
755 {
756 	int nmi_wd_enabled = apic_lvt_nmi_mode(apic_get_reg(apic, APIC_LVT0));
757 
758 	if (apic_lvt_nmi_mode(lvt0_val)) {
759 		if (!nmi_wd_enabled) {
760 			apic_debug("Receive NMI setting on APIC_LVT0 "
761 				   "for cpu %d\n", apic->vcpu->vcpu_id);
762 			apic->vcpu->kvm->arch.vapics_in_nmi_mode++;
763 		}
764 	} else if (nmi_wd_enabled)
765 		apic->vcpu->kvm->arch.vapics_in_nmi_mode--;
766 }
767 
768 static int apic_reg_write(struct kvm_lapic *apic, u32 reg, u32 val)
769 {
770 	int ret = 0;
771 
772 	trace_kvm_apic_write(reg, val);
773 
774 	switch (reg) {
775 	case APIC_ID:		/* Local APIC ID */
776 		if (!apic_x2apic_mode(apic))
777 			apic_set_reg(apic, APIC_ID, val);
778 		else
779 			ret = 1;
780 		break;
781 
782 	case APIC_TASKPRI:
783 		report_tpr_access(apic, true);
784 		apic_set_tpr(apic, val & 0xff);
785 		break;
786 
787 	case APIC_EOI:
788 		apic_set_eoi(apic);
789 		break;
790 
791 	case APIC_LDR:
792 		if (!apic_x2apic_mode(apic))
793 			apic_set_reg(apic, APIC_LDR, val & APIC_LDR_MASK);
794 		else
795 			ret = 1;
796 		break;
797 
798 	case APIC_DFR:
799 		if (!apic_x2apic_mode(apic))
800 			apic_set_reg(apic, APIC_DFR, val | 0x0FFFFFFF);
801 		else
802 			ret = 1;
803 		break;
804 
805 	case APIC_SPIV: {
806 		u32 mask = 0x3ff;
807 		if (apic_get_reg(apic, APIC_LVR) & APIC_LVR_DIRECTED_EOI)
808 			mask |= APIC_SPIV_DIRECTED_EOI;
809 		apic_set_reg(apic, APIC_SPIV, val & mask);
810 		if (!(val & APIC_SPIV_APIC_ENABLED)) {
811 			int i;
812 			u32 lvt_val;
813 
814 			for (i = 0; i < APIC_LVT_NUM; i++) {
815 				lvt_val = apic_get_reg(apic,
816 						       APIC_LVTT + 0x10 * i);
817 				apic_set_reg(apic, APIC_LVTT + 0x10 * i,
818 					     lvt_val | APIC_LVT_MASKED);
819 			}
820 			atomic_set(&apic->lapic_timer.pending, 0);
821 
822 		}
823 		break;
824 	}
825 	case APIC_ICR:
826 		/* No delay here, so we always clear the pending bit */
827 		apic_set_reg(apic, APIC_ICR, val & ~(1 << 12));
828 		apic_send_ipi(apic);
829 		break;
830 
831 	case APIC_ICR2:
832 		if (!apic_x2apic_mode(apic))
833 			val &= 0xff000000;
834 		apic_set_reg(apic, APIC_ICR2, val);
835 		break;
836 
837 	case APIC_LVT0:
838 		apic_manage_nmi_watchdog(apic, val);
839 	case APIC_LVTTHMR:
840 	case APIC_LVTPC:
841 	case APIC_LVT1:
842 	case APIC_LVTERR:
843 		/* TODO: Check vector */
844 		if (!apic_sw_enabled(apic))
845 			val |= APIC_LVT_MASKED;
846 
847 		val &= apic_lvt_mask[(reg - APIC_LVTT) >> 4];
848 		apic_set_reg(apic, reg, val);
849 
850 		break;
851 
852 	case APIC_LVTT:
853 		if ((apic_get_reg(apic, APIC_LVTT) &
854 		    apic->lapic_timer.timer_mode_mask) !=
855 		   (val & apic->lapic_timer.timer_mode_mask))
856 			hrtimer_cancel(&apic->lapic_timer.timer);
857 
858 		if (!apic_sw_enabled(apic))
859 			val |= APIC_LVT_MASKED;
860 		val &= (apic_lvt_mask[0] | apic->lapic_timer.timer_mode_mask);
861 		apic_set_reg(apic, APIC_LVTT, val);
862 		break;
863 
864 	case APIC_TMICT:
865 		if (apic_lvtt_tscdeadline(apic))
866 			break;
867 
868 		hrtimer_cancel(&apic->lapic_timer.timer);
869 		apic_set_reg(apic, APIC_TMICT, val);
870 		start_apic_timer(apic);
871 		break;
872 
873 	case APIC_TDCR:
874 		if (val & 4)
875 			apic_debug("KVM_WRITE:TDCR %x\n", val);
876 		apic_set_reg(apic, APIC_TDCR, val);
877 		update_divide_count(apic);
878 		break;
879 
880 	case APIC_ESR:
881 		if (apic_x2apic_mode(apic) && val != 0) {
882 			apic_debug("KVM_WRITE:ESR not zero %x\n", val);
883 			ret = 1;
884 		}
885 		break;
886 
887 	case APIC_SELF_IPI:
888 		if (apic_x2apic_mode(apic)) {
889 			apic_reg_write(apic, APIC_ICR, 0x40000 | (val & 0xff));
890 		} else
891 			ret = 1;
892 		break;
893 	default:
894 		ret = 1;
895 		break;
896 	}
897 	if (ret)
898 		apic_debug("Local APIC Write to read-only register %x\n", reg);
899 	return ret;
900 }
901 
902 static int apic_mmio_write(struct kvm_io_device *this,
903 			    gpa_t address, int len, const void *data)
904 {
905 	struct kvm_lapic *apic = to_lapic(this);
906 	unsigned int offset = address - apic->base_address;
907 	u32 val;
908 
909 	if (!apic_mmio_in_range(apic, address))
910 		return -EOPNOTSUPP;
911 
912 	/*
913 	 * APIC register must be aligned on 128-bits boundary.
914 	 * 32/64/128 bits registers must be accessed thru 32 bits.
915 	 * Refer SDM 8.4.1
916 	 */
917 	if (len != 4 || (offset & 0xf)) {
918 		/* Don't shout loud, $infamous_os would cause only noise. */
919 		apic_debug("apic write: bad size=%d %lx\n", len, (long)address);
920 		return 0;
921 	}
922 
923 	val = *(u32*)data;
924 
925 	/* too common printing */
926 	if (offset != APIC_EOI)
927 		apic_debug("%s: offset 0x%x with length 0x%x, and value is "
928 			   "0x%x\n", __func__, offset, len, val);
929 
930 	apic_reg_write(apic, offset & 0xff0, val);
931 
932 	return 0;
933 }
934 
935 void kvm_lapic_set_eoi(struct kvm_vcpu *vcpu)
936 {
937 	struct kvm_lapic *apic = vcpu->arch.apic;
938 
939 	if (apic)
940 		apic_reg_write(vcpu->arch.apic, APIC_EOI, 0);
941 }
942 EXPORT_SYMBOL_GPL(kvm_lapic_set_eoi);
943 
944 void kvm_free_lapic(struct kvm_vcpu *vcpu)
945 {
946 	if (!vcpu->arch.apic)
947 		return;
948 
949 	hrtimer_cancel(&vcpu->arch.apic->lapic_timer.timer);
950 
951 	if (vcpu->arch.apic->regs)
952 		free_page((unsigned long)vcpu->arch.apic->regs);
953 
954 	kfree(vcpu->arch.apic);
955 }
956 
957 /*
958  *----------------------------------------------------------------------
959  * LAPIC interface
960  *----------------------------------------------------------------------
961  */
962 
963 u64 kvm_get_lapic_tscdeadline_msr(struct kvm_vcpu *vcpu)
964 {
965 	struct kvm_lapic *apic = vcpu->arch.apic;
966 	if (!apic)
967 		return 0;
968 
969 	if (apic_lvtt_oneshot(apic) || apic_lvtt_period(apic))
970 		return 0;
971 
972 	return apic->lapic_timer.tscdeadline;
973 }
974 
975 void kvm_set_lapic_tscdeadline_msr(struct kvm_vcpu *vcpu, u64 data)
976 {
977 	struct kvm_lapic *apic = vcpu->arch.apic;
978 	if (!apic)
979 		return;
980 
981 	if (apic_lvtt_oneshot(apic) || apic_lvtt_period(apic))
982 		return;
983 
984 	hrtimer_cancel(&apic->lapic_timer.timer);
985 	apic->lapic_timer.tscdeadline = data;
986 	start_apic_timer(apic);
987 }
988 
989 void kvm_lapic_set_tpr(struct kvm_vcpu *vcpu, unsigned long cr8)
990 {
991 	struct kvm_lapic *apic = vcpu->arch.apic;
992 
993 	if (!apic)
994 		return;
995 	apic_set_tpr(apic, ((cr8 & 0x0f) << 4)
996 		     | (apic_get_reg(apic, APIC_TASKPRI) & 4));
997 }
998 
999 u64 kvm_lapic_get_cr8(struct kvm_vcpu *vcpu)
1000 {
1001 	struct kvm_lapic *apic = vcpu->arch.apic;
1002 	u64 tpr;
1003 
1004 	if (!apic)
1005 		return 0;
1006 	tpr = (u64) apic_get_reg(apic, APIC_TASKPRI);
1007 
1008 	return (tpr & 0xf0) >> 4;
1009 }
1010 
1011 void kvm_lapic_set_base(struct kvm_vcpu *vcpu, u64 value)
1012 {
1013 	struct kvm_lapic *apic = vcpu->arch.apic;
1014 
1015 	if (!apic) {
1016 		value |= MSR_IA32_APICBASE_BSP;
1017 		vcpu->arch.apic_base = value;
1018 		return;
1019 	}
1020 
1021 	if (!kvm_vcpu_is_bsp(apic->vcpu))
1022 		value &= ~MSR_IA32_APICBASE_BSP;
1023 
1024 	vcpu->arch.apic_base = value;
1025 	if (apic_x2apic_mode(apic)) {
1026 		u32 id = kvm_apic_id(apic);
1027 		u32 ldr = ((id & ~0xf) << 16) | (1 << (id & 0xf));
1028 		apic_set_reg(apic, APIC_LDR, ldr);
1029 	}
1030 	apic->base_address = apic->vcpu->arch.apic_base &
1031 			     MSR_IA32_APICBASE_BASE;
1032 
1033 	/* with FSB delivery interrupt, we can restart APIC functionality */
1034 	apic_debug("apic base msr is 0x%016" PRIx64 ", and base address is "
1035 		   "0x%lx.\n", apic->vcpu->arch.apic_base, apic->base_address);
1036 
1037 }
1038 
1039 void kvm_lapic_reset(struct kvm_vcpu *vcpu)
1040 {
1041 	struct kvm_lapic *apic;
1042 	int i;
1043 
1044 	apic_debug("%s\n", __func__);
1045 
1046 	ASSERT(vcpu);
1047 	apic = vcpu->arch.apic;
1048 	ASSERT(apic != NULL);
1049 
1050 	/* Stop the timer in case it's a reset to an active apic */
1051 	hrtimer_cancel(&apic->lapic_timer.timer);
1052 
1053 	apic_set_reg(apic, APIC_ID, vcpu->vcpu_id << 24);
1054 	kvm_apic_set_version(apic->vcpu);
1055 
1056 	for (i = 0; i < APIC_LVT_NUM; i++)
1057 		apic_set_reg(apic, APIC_LVTT + 0x10 * i, APIC_LVT_MASKED);
1058 	apic_set_reg(apic, APIC_LVT0,
1059 		     SET_APIC_DELIVERY_MODE(0, APIC_MODE_EXTINT));
1060 
1061 	apic_set_reg(apic, APIC_DFR, 0xffffffffU);
1062 	apic_set_reg(apic, APIC_SPIV, 0xff);
1063 	apic_set_reg(apic, APIC_TASKPRI, 0);
1064 	apic_set_reg(apic, APIC_LDR, 0);
1065 	apic_set_reg(apic, APIC_ESR, 0);
1066 	apic_set_reg(apic, APIC_ICR, 0);
1067 	apic_set_reg(apic, APIC_ICR2, 0);
1068 	apic_set_reg(apic, APIC_TDCR, 0);
1069 	apic_set_reg(apic, APIC_TMICT, 0);
1070 	for (i = 0; i < 8; i++) {
1071 		apic_set_reg(apic, APIC_IRR + 0x10 * i, 0);
1072 		apic_set_reg(apic, APIC_ISR + 0x10 * i, 0);
1073 		apic_set_reg(apic, APIC_TMR + 0x10 * i, 0);
1074 	}
1075 	apic->irr_pending = false;
1076 	update_divide_count(apic);
1077 	atomic_set(&apic->lapic_timer.pending, 0);
1078 	if (kvm_vcpu_is_bsp(vcpu))
1079 		vcpu->arch.apic_base |= MSR_IA32_APICBASE_BSP;
1080 	apic_update_ppr(apic);
1081 
1082 	vcpu->arch.apic_arb_prio = 0;
1083 
1084 	apic_debug(KERN_INFO "%s: vcpu=%p, id=%d, base_msr="
1085 		   "0x%016" PRIx64 ", base_address=0x%0lx.\n", __func__,
1086 		   vcpu, kvm_apic_id(apic),
1087 		   vcpu->arch.apic_base, apic->base_address);
1088 }
1089 
1090 bool kvm_apic_present(struct kvm_vcpu *vcpu)
1091 {
1092 	return vcpu->arch.apic && apic_hw_enabled(vcpu->arch.apic);
1093 }
1094 
1095 int kvm_lapic_enabled(struct kvm_vcpu *vcpu)
1096 {
1097 	return kvm_apic_present(vcpu) && apic_sw_enabled(vcpu->arch.apic);
1098 }
1099 
1100 /*
1101  *----------------------------------------------------------------------
1102  * timer interface
1103  *----------------------------------------------------------------------
1104  */
1105 
1106 static bool lapic_is_periodic(struct kvm_timer *ktimer)
1107 {
1108 	struct kvm_lapic *apic = container_of(ktimer, struct kvm_lapic,
1109 					      lapic_timer);
1110 	return apic_lvtt_period(apic);
1111 }
1112 
1113 int apic_has_pending_timer(struct kvm_vcpu *vcpu)
1114 {
1115 	struct kvm_lapic *lapic = vcpu->arch.apic;
1116 
1117 	if (lapic && apic_enabled(lapic) && apic_lvt_enabled(lapic, APIC_LVTT))
1118 		return atomic_read(&lapic->lapic_timer.pending);
1119 
1120 	return 0;
1121 }
1122 
1123 static int kvm_apic_local_deliver(struct kvm_lapic *apic, int lvt_type)
1124 {
1125 	u32 reg = apic_get_reg(apic, lvt_type);
1126 	int vector, mode, trig_mode;
1127 
1128 	if (apic_hw_enabled(apic) && !(reg & APIC_LVT_MASKED)) {
1129 		vector = reg & APIC_VECTOR_MASK;
1130 		mode = reg & APIC_MODE_MASK;
1131 		trig_mode = reg & APIC_LVT_LEVEL_TRIGGER;
1132 		return __apic_accept_irq(apic, mode, vector, 1, trig_mode);
1133 	}
1134 	return 0;
1135 }
1136 
1137 void kvm_apic_nmi_wd_deliver(struct kvm_vcpu *vcpu)
1138 {
1139 	struct kvm_lapic *apic = vcpu->arch.apic;
1140 
1141 	if (apic)
1142 		kvm_apic_local_deliver(apic, APIC_LVT0);
1143 }
1144 
1145 static struct kvm_timer_ops lapic_timer_ops = {
1146 	.is_periodic = lapic_is_periodic,
1147 };
1148 
1149 static const struct kvm_io_device_ops apic_mmio_ops = {
1150 	.read     = apic_mmio_read,
1151 	.write    = apic_mmio_write,
1152 };
1153 
1154 int kvm_create_lapic(struct kvm_vcpu *vcpu)
1155 {
1156 	struct kvm_lapic *apic;
1157 
1158 	ASSERT(vcpu != NULL);
1159 	apic_debug("apic_init %d\n", vcpu->vcpu_id);
1160 
1161 	apic = kzalloc(sizeof(*apic), GFP_KERNEL);
1162 	if (!apic)
1163 		goto nomem;
1164 
1165 	vcpu->arch.apic = apic;
1166 
1167 	apic->regs = (void *)get_zeroed_page(GFP_KERNEL);
1168 	if (!apic->regs) {
1169 		printk(KERN_ERR "malloc apic regs error for vcpu %x\n",
1170 		       vcpu->vcpu_id);
1171 		goto nomem_free_apic;
1172 	}
1173 	apic->vcpu = vcpu;
1174 
1175 	hrtimer_init(&apic->lapic_timer.timer, CLOCK_MONOTONIC,
1176 		     HRTIMER_MODE_ABS);
1177 	apic->lapic_timer.timer.function = kvm_timer_fn;
1178 	apic->lapic_timer.t_ops = &lapic_timer_ops;
1179 	apic->lapic_timer.kvm = vcpu->kvm;
1180 	apic->lapic_timer.vcpu = vcpu;
1181 
1182 	apic->base_address = APIC_DEFAULT_PHYS_BASE;
1183 	vcpu->arch.apic_base = APIC_DEFAULT_PHYS_BASE;
1184 
1185 	kvm_lapic_reset(vcpu);
1186 	kvm_iodevice_init(&apic->dev, &apic_mmio_ops);
1187 
1188 	return 0;
1189 nomem_free_apic:
1190 	kfree(apic);
1191 nomem:
1192 	return -ENOMEM;
1193 }
1194 
1195 int kvm_apic_has_interrupt(struct kvm_vcpu *vcpu)
1196 {
1197 	struct kvm_lapic *apic = vcpu->arch.apic;
1198 	int highest_irr;
1199 
1200 	if (!apic || !apic_enabled(apic))
1201 		return -1;
1202 
1203 	apic_update_ppr(apic);
1204 	highest_irr = apic_find_highest_irr(apic);
1205 	if ((highest_irr == -1) ||
1206 	    ((highest_irr & 0xF0) <= apic_get_reg(apic, APIC_PROCPRI)))
1207 		return -1;
1208 	return highest_irr;
1209 }
1210 
1211 int kvm_apic_accept_pic_intr(struct kvm_vcpu *vcpu)
1212 {
1213 	u32 lvt0 = apic_get_reg(vcpu->arch.apic, APIC_LVT0);
1214 	int r = 0;
1215 
1216 	if (!apic_hw_enabled(vcpu->arch.apic))
1217 		r = 1;
1218 	if ((lvt0 & APIC_LVT_MASKED) == 0 &&
1219 	    GET_APIC_DELIVERY_MODE(lvt0) == APIC_MODE_EXTINT)
1220 		r = 1;
1221 	return r;
1222 }
1223 
1224 void kvm_inject_apic_timer_irqs(struct kvm_vcpu *vcpu)
1225 {
1226 	struct kvm_lapic *apic = vcpu->arch.apic;
1227 
1228 	if (apic && atomic_read(&apic->lapic_timer.pending) > 0) {
1229 		if (kvm_apic_local_deliver(apic, APIC_LVTT))
1230 			atomic_dec(&apic->lapic_timer.pending);
1231 	}
1232 }
1233 
1234 int kvm_get_apic_interrupt(struct kvm_vcpu *vcpu)
1235 {
1236 	int vector = kvm_apic_has_interrupt(vcpu);
1237 	struct kvm_lapic *apic = vcpu->arch.apic;
1238 
1239 	if (vector == -1)
1240 		return -1;
1241 
1242 	apic_set_vector(vector, apic->regs + APIC_ISR);
1243 	apic_update_ppr(apic);
1244 	apic_clear_irr(vector, apic);
1245 	return vector;
1246 }
1247 
1248 void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu)
1249 {
1250 	struct kvm_lapic *apic = vcpu->arch.apic;
1251 
1252 	apic->base_address = vcpu->arch.apic_base &
1253 			     MSR_IA32_APICBASE_BASE;
1254 	kvm_apic_set_version(vcpu);
1255 
1256 	apic_update_ppr(apic);
1257 	hrtimer_cancel(&apic->lapic_timer.timer);
1258 	update_divide_count(apic);
1259 	start_apic_timer(apic);
1260 	apic->irr_pending = true;
1261 	kvm_make_request(KVM_REQ_EVENT, vcpu);
1262 }
1263 
1264 void __kvm_migrate_apic_timer(struct kvm_vcpu *vcpu)
1265 {
1266 	struct kvm_lapic *apic = vcpu->arch.apic;
1267 	struct hrtimer *timer;
1268 
1269 	if (!apic)
1270 		return;
1271 
1272 	timer = &apic->lapic_timer.timer;
1273 	if (hrtimer_cancel(timer))
1274 		hrtimer_start_expires(timer, HRTIMER_MODE_ABS);
1275 }
1276 
1277 void kvm_lapic_sync_from_vapic(struct kvm_vcpu *vcpu)
1278 {
1279 	u32 data;
1280 	void *vapic;
1281 
1282 	if (!irqchip_in_kernel(vcpu->kvm) || !vcpu->arch.apic->vapic_addr)
1283 		return;
1284 
1285 	vapic = kmap_atomic(vcpu->arch.apic->vapic_page, KM_USER0);
1286 	data = *(u32 *)(vapic + offset_in_page(vcpu->arch.apic->vapic_addr));
1287 	kunmap_atomic(vapic, KM_USER0);
1288 
1289 	apic_set_tpr(vcpu->arch.apic, data & 0xff);
1290 }
1291 
1292 void kvm_lapic_sync_to_vapic(struct kvm_vcpu *vcpu)
1293 {
1294 	u32 data, tpr;
1295 	int max_irr, max_isr;
1296 	struct kvm_lapic *apic;
1297 	void *vapic;
1298 
1299 	if (!irqchip_in_kernel(vcpu->kvm) || !vcpu->arch.apic->vapic_addr)
1300 		return;
1301 
1302 	apic = vcpu->arch.apic;
1303 	tpr = apic_get_reg(apic, APIC_TASKPRI) & 0xff;
1304 	max_irr = apic_find_highest_irr(apic);
1305 	if (max_irr < 0)
1306 		max_irr = 0;
1307 	max_isr = apic_find_highest_isr(apic);
1308 	if (max_isr < 0)
1309 		max_isr = 0;
1310 	data = (tpr & 0xff) | ((max_isr & 0xf0) << 8) | (max_irr << 24);
1311 
1312 	vapic = kmap_atomic(vcpu->arch.apic->vapic_page, KM_USER0);
1313 	*(u32 *)(vapic + offset_in_page(vcpu->arch.apic->vapic_addr)) = data;
1314 	kunmap_atomic(vapic, KM_USER0);
1315 }
1316 
1317 void kvm_lapic_set_vapic_addr(struct kvm_vcpu *vcpu, gpa_t vapic_addr)
1318 {
1319 	if (!irqchip_in_kernel(vcpu->kvm))
1320 		return;
1321 
1322 	vcpu->arch.apic->vapic_addr = vapic_addr;
1323 }
1324 
1325 int kvm_x2apic_msr_write(struct kvm_vcpu *vcpu, u32 msr, u64 data)
1326 {
1327 	struct kvm_lapic *apic = vcpu->arch.apic;
1328 	u32 reg = (msr - APIC_BASE_MSR) << 4;
1329 
1330 	if (!irqchip_in_kernel(vcpu->kvm) || !apic_x2apic_mode(apic))
1331 		return 1;
1332 
1333 	/* if this is ICR write vector before command */
1334 	if (msr == 0x830)
1335 		apic_reg_write(apic, APIC_ICR2, (u32)(data >> 32));
1336 	return apic_reg_write(apic, reg, (u32)data);
1337 }
1338 
1339 int kvm_x2apic_msr_read(struct kvm_vcpu *vcpu, u32 msr, u64 *data)
1340 {
1341 	struct kvm_lapic *apic = vcpu->arch.apic;
1342 	u32 reg = (msr - APIC_BASE_MSR) << 4, low, high = 0;
1343 
1344 	if (!irqchip_in_kernel(vcpu->kvm) || !apic_x2apic_mode(apic))
1345 		return 1;
1346 
1347 	if (apic_reg_read(apic, reg, 4, &low))
1348 		return 1;
1349 	if (msr == 0x830)
1350 		apic_reg_read(apic, APIC_ICR2, 4, &high);
1351 
1352 	*data = (((u64)high) << 32) | low;
1353 
1354 	return 0;
1355 }
1356 
1357 int kvm_hv_vapic_msr_write(struct kvm_vcpu *vcpu, u32 reg, u64 data)
1358 {
1359 	struct kvm_lapic *apic = vcpu->arch.apic;
1360 
1361 	if (!irqchip_in_kernel(vcpu->kvm))
1362 		return 1;
1363 
1364 	/* if this is ICR write vector before command */
1365 	if (reg == APIC_ICR)
1366 		apic_reg_write(apic, APIC_ICR2, (u32)(data >> 32));
1367 	return apic_reg_write(apic, reg, (u32)data);
1368 }
1369 
1370 int kvm_hv_vapic_msr_read(struct kvm_vcpu *vcpu, u32 reg, u64 *data)
1371 {
1372 	struct kvm_lapic *apic = vcpu->arch.apic;
1373 	u32 low, high = 0;
1374 
1375 	if (!irqchip_in_kernel(vcpu->kvm))
1376 		return 1;
1377 
1378 	if (apic_reg_read(apic, reg, 4, &low))
1379 		return 1;
1380 	if (reg == APIC_ICR)
1381 		apic_reg_read(apic, APIC_ICR2, 4, &high);
1382 
1383 	*data = (((u64)high) << 32) | low;
1384 
1385 	return 0;
1386 }
1387