xref: /linux/arch/x86/kvm/lapic.c (revision 26b0d14106954ae46d2f4f7eec3481828a210f7d)
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 #include "cpuid.h"
42 
43 #ifndef CONFIG_X86_64
44 #define mod_64(x, y) ((x) - (y) * div64_u64(x, y))
45 #else
46 #define mod_64(x, y) ((x) % (y))
47 #endif
48 
49 #define PRId64 "d"
50 #define PRIx64 "llx"
51 #define PRIu64 "u"
52 #define PRIo64 "o"
53 
54 #define APIC_BUS_CYCLE_NS 1
55 
56 /* #define apic_debug(fmt,arg...) printk(KERN_WARNING fmt,##arg) */
57 #define apic_debug(fmt, arg...)
58 
59 #define APIC_LVT_NUM			6
60 /* 14 is the version for Xeon and Pentium 8.4.8*/
61 #define APIC_VERSION			(0x14UL | ((APIC_LVT_NUM - 1) << 16))
62 #define LAPIC_MMIO_LENGTH		(1 << 12)
63 /* followed define is not in apicdef.h */
64 #define APIC_SHORT_MASK			0xc0000
65 #define APIC_DEST_NOSHORT		0x0
66 #define APIC_DEST_MASK			0x800
67 #define MAX_APIC_VECTOR			256
68 
69 #define VEC_POS(v) ((v) & (32 - 1))
70 #define REG_POS(v) (((v) >> 5) << 4)
71 
72 static unsigned int min_timer_period_us = 500;
73 module_param(min_timer_period_us, uint, S_IRUGO | S_IWUSR);
74 
75 static inline u32 apic_get_reg(struct kvm_lapic *apic, int reg_off)
76 {
77 	return *((u32 *) (apic->regs + reg_off));
78 }
79 
80 static inline void apic_set_reg(struct kvm_lapic *apic, int reg_off, u32 val)
81 {
82 	*((u32 *) (apic->regs + reg_off)) = val;
83 }
84 
85 static inline int apic_test_and_set_vector(int vec, void *bitmap)
86 {
87 	return test_and_set_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
88 }
89 
90 static inline int apic_test_and_clear_vector(int vec, void *bitmap)
91 {
92 	return test_and_clear_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
93 }
94 
95 static inline int apic_test_vector(int vec, void *bitmap)
96 {
97 	return test_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
98 }
99 
100 static inline void apic_set_vector(int vec, void *bitmap)
101 {
102 	set_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
103 }
104 
105 static inline void apic_clear_vector(int vec, void *bitmap)
106 {
107 	clear_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
108 }
109 
110 static inline int apic_hw_enabled(struct kvm_lapic *apic)
111 {
112 	return (apic)->vcpu->arch.apic_base & MSR_IA32_APICBASE_ENABLE;
113 }
114 
115 static inline int  apic_sw_enabled(struct kvm_lapic *apic)
116 {
117 	return apic_get_reg(apic, APIC_SPIV) & APIC_SPIV_APIC_ENABLED;
118 }
119 
120 static inline int apic_enabled(struct kvm_lapic *apic)
121 {
122 	return apic_sw_enabled(apic) &&	apic_hw_enabled(apic);
123 }
124 
125 #define LVT_MASK	\
126 	(APIC_LVT_MASKED | APIC_SEND_PENDING | APIC_VECTOR_MASK)
127 
128 #define LINT_MASK	\
129 	(LVT_MASK | APIC_MODE_MASK | APIC_INPUT_POLARITY | \
130 	 APIC_LVT_REMOTE_IRR | APIC_LVT_LEVEL_TRIGGER)
131 
132 static inline int kvm_apic_id(struct kvm_lapic *apic)
133 {
134 	return (apic_get_reg(apic, APIC_ID) >> 24) & 0xff;
135 }
136 
137 static inline int apic_lvt_enabled(struct kvm_lapic *apic, int lvt_type)
138 {
139 	return !(apic_get_reg(apic, lvt_type) & APIC_LVT_MASKED);
140 }
141 
142 static inline int apic_lvt_vector(struct kvm_lapic *apic, int lvt_type)
143 {
144 	return apic_get_reg(apic, lvt_type) & APIC_VECTOR_MASK;
145 }
146 
147 static inline int apic_lvtt_oneshot(struct kvm_lapic *apic)
148 {
149 	return ((apic_get_reg(apic, APIC_LVTT) &
150 		apic->lapic_timer.timer_mode_mask) == APIC_LVT_TIMER_ONESHOT);
151 }
152 
153 static inline int apic_lvtt_period(struct kvm_lapic *apic)
154 {
155 	return ((apic_get_reg(apic, APIC_LVTT) &
156 		apic->lapic_timer.timer_mode_mask) == APIC_LVT_TIMER_PERIODIC);
157 }
158 
159 static inline int apic_lvtt_tscdeadline(struct kvm_lapic *apic)
160 {
161 	return ((apic_get_reg(apic, APIC_LVTT) &
162 		apic->lapic_timer.timer_mode_mask) ==
163 			APIC_LVT_TIMER_TSCDEADLINE);
164 }
165 
166 static inline int apic_lvt_nmi_mode(u32 lvt_val)
167 {
168 	return (lvt_val & (APIC_MODE_MASK | APIC_LVT_MASKED)) == APIC_DM_NMI;
169 }
170 
171 void kvm_apic_set_version(struct kvm_vcpu *vcpu)
172 {
173 	struct kvm_lapic *apic = vcpu->arch.apic;
174 	struct kvm_cpuid_entry2 *feat;
175 	u32 v = APIC_VERSION;
176 
177 	if (!irqchip_in_kernel(vcpu->kvm))
178 		return;
179 
180 	feat = kvm_find_cpuid_entry(apic->vcpu, 0x1, 0);
181 	if (feat && (feat->ecx & (1 << (X86_FEATURE_X2APIC & 31))))
182 		v |= APIC_LVR_DIRECTED_EOI;
183 	apic_set_reg(apic, APIC_LVR, v);
184 }
185 
186 static inline int apic_x2apic_mode(struct kvm_lapic *apic)
187 {
188 	return apic->vcpu->arch.apic_base & X2APIC_ENABLE;
189 }
190 
191 static unsigned int apic_lvt_mask[APIC_LVT_NUM] = {
192 	LVT_MASK ,      /* part LVTT mask, timer mode mask added at runtime */
193 	LVT_MASK | APIC_MODE_MASK,	/* LVTTHMR */
194 	LVT_MASK | APIC_MODE_MASK,	/* LVTPC */
195 	LINT_MASK, LINT_MASK,	/* LVT0-1 */
196 	LVT_MASK		/* LVTERR */
197 };
198 
199 static int find_highest_vector(void *bitmap)
200 {
201 	u32 *word = bitmap;
202 	int word_offset = MAX_APIC_VECTOR >> 5;
203 
204 	while ((word_offset != 0) && (word[(--word_offset) << 2] == 0))
205 		continue;
206 
207 	if (likely(!word_offset && !word[0]))
208 		return -1;
209 	else
210 		return fls(word[word_offset << 2]) - 1 + (word_offset << 5);
211 }
212 
213 static inline int apic_test_and_set_irr(int vec, struct kvm_lapic *apic)
214 {
215 	apic->irr_pending = true;
216 	return apic_test_and_set_vector(vec, apic->regs + APIC_IRR);
217 }
218 
219 static inline int apic_search_irr(struct kvm_lapic *apic)
220 {
221 	return find_highest_vector(apic->regs + APIC_IRR);
222 }
223 
224 static inline int apic_find_highest_irr(struct kvm_lapic *apic)
225 {
226 	int result;
227 
228 	if (!apic->irr_pending)
229 		return -1;
230 
231 	result = apic_search_irr(apic);
232 	ASSERT(result == -1 || result >= 16);
233 
234 	return result;
235 }
236 
237 static inline void apic_clear_irr(int vec, struct kvm_lapic *apic)
238 {
239 	apic->irr_pending = false;
240 	apic_clear_vector(vec, apic->regs + APIC_IRR);
241 	if (apic_search_irr(apic) != -1)
242 		apic->irr_pending = true;
243 }
244 
245 int kvm_lapic_find_highest_irr(struct kvm_vcpu *vcpu)
246 {
247 	struct kvm_lapic *apic = vcpu->arch.apic;
248 	int highest_irr;
249 
250 	/* This may race with setting of irr in __apic_accept_irq() and
251 	 * value returned may be wrong, but kvm_vcpu_kick() in __apic_accept_irq
252 	 * will cause vmexit immediately and the value will be recalculated
253 	 * on the next vmentry.
254 	 */
255 	if (!apic)
256 		return 0;
257 	highest_irr = apic_find_highest_irr(apic);
258 
259 	return highest_irr;
260 }
261 
262 static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
263 			     int vector, int level, int trig_mode);
264 
265 int kvm_apic_set_irq(struct kvm_vcpu *vcpu, struct kvm_lapic_irq *irq)
266 {
267 	struct kvm_lapic *apic = vcpu->arch.apic;
268 
269 	return __apic_accept_irq(apic, irq->delivery_mode, irq->vector,
270 			irq->level, irq->trig_mode);
271 }
272 
273 static inline int apic_find_highest_isr(struct kvm_lapic *apic)
274 {
275 	int result;
276 
277 	result = find_highest_vector(apic->regs + APIC_ISR);
278 	ASSERT(result == -1 || result >= 16);
279 
280 	return result;
281 }
282 
283 static void apic_update_ppr(struct kvm_lapic *apic)
284 {
285 	u32 tpr, isrv, ppr, old_ppr;
286 	int isr;
287 
288 	old_ppr = apic_get_reg(apic, APIC_PROCPRI);
289 	tpr = apic_get_reg(apic, APIC_TASKPRI);
290 	isr = apic_find_highest_isr(apic);
291 	isrv = (isr != -1) ? isr : 0;
292 
293 	if ((tpr & 0xf0) >= (isrv & 0xf0))
294 		ppr = tpr & 0xff;
295 	else
296 		ppr = isrv & 0xf0;
297 
298 	apic_debug("vlapic %p, ppr 0x%x, isr 0x%x, isrv 0x%x",
299 		   apic, ppr, isr, isrv);
300 
301 	if (old_ppr != ppr) {
302 		apic_set_reg(apic, APIC_PROCPRI, ppr);
303 		if (ppr < old_ppr)
304 			kvm_make_request(KVM_REQ_EVENT, apic->vcpu);
305 	}
306 }
307 
308 static void apic_set_tpr(struct kvm_lapic *apic, u32 tpr)
309 {
310 	apic_set_reg(apic, APIC_TASKPRI, tpr);
311 	apic_update_ppr(apic);
312 }
313 
314 int kvm_apic_match_physical_addr(struct kvm_lapic *apic, u16 dest)
315 {
316 	return dest == 0xff || kvm_apic_id(apic) == dest;
317 }
318 
319 int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
320 {
321 	int result = 0;
322 	u32 logical_id;
323 
324 	if (apic_x2apic_mode(apic)) {
325 		logical_id = apic_get_reg(apic, APIC_LDR);
326 		return logical_id & mda;
327 	}
328 
329 	logical_id = GET_APIC_LOGICAL_ID(apic_get_reg(apic, APIC_LDR));
330 
331 	switch (apic_get_reg(apic, APIC_DFR)) {
332 	case APIC_DFR_FLAT:
333 		if (logical_id & mda)
334 			result = 1;
335 		break;
336 	case APIC_DFR_CLUSTER:
337 		if (((logical_id >> 4) == (mda >> 0x4))
338 		    && (logical_id & mda & 0xf))
339 			result = 1;
340 		break;
341 	default:
342 		apic_debug("Bad DFR vcpu %d: %08x\n",
343 			   apic->vcpu->vcpu_id, apic_get_reg(apic, APIC_DFR));
344 		break;
345 	}
346 
347 	return result;
348 }
349 
350 int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
351 			   int short_hand, int dest, int dest_mode)
352 {
353 	int result = 0;
354 	struct kvm_lapic *target = vcpu->arch.apic;
355 
356 	apic_debug("target %p, source %p, dest 0x%x, "
357 		   "dest_mode 0x%x, short_hand 0x%x\n",
358 		   target, source, dest, dest_mode, short_hand);
359 
360 	ASSERT(target);
361 	switch (short_hand) {
362 	case APIC_DEST_NOSHORT:
363 		if (dest_mode == 0)
364 			/* Physical mode. */
365 			result = kvm_apic_match_physical_addr(target, dest);
366 		else
367 			/* Logical mode. */
368 			result = kvm_apic_match_logical_addr(target, dest);
369 		break;
370 	case APIC_DEST_SELF:
371 		result = (target == source);
372 		break;
373 	case APIC_DEST_ALLINC:
374 		result = 1;
375 		break;
376 	case APIC_DEST_ALLBUT:
377 		result = (target != source);
378 		break;
379 	default:
380 		apic_debug("kvm: apic: Bad dest shorthand value %x\n",
381 			   short_hand);
382 		break;
383 	}
384 
385 	return result;
386 }
387 
388 /*
389  * Add a pending IRQ into lapic.
390  * Return 1 if successfully added and 0 if discarded.
391  */
392 static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
393 			     int vector, int level, int trig_mode)
394 {
395 	int result = 0;
396 	struct kvm_vcpu *vcpu = apic->vcpu;
397 
398 	switch (delivery_mode) {
399 	case APIC_DM_LOWEST:
400 		vcpu->arch.apic_arb_prio++;
401 	case APIC_DM_FIXED:
402 		/* FIXME add logic for vcpu on reset */
403 		if (unlikely(!apic_enabled(apic)))
404 			break;
405 
406 		if (trig_mode) {
407 			apic_debug("level trig mode for vector %d", vector);
408 			apic_set_vector(vector, apic->regs + APIC_TMR);
409 		} else
410 			apic_clear_vector(vector, apic->regs + APIC_TMR);
411 
412 		result = !apic_test_and_set_irr(vector, apic);
413 		trace_kvm_apic_accept_irq(vcpu->vcpu_id, delivery_mode,
414 					  trig_mode, vector, !result);
415 		if (!result) {
416 			if (trig_mode)
417 				apic_debug("level trig mode repeatedly for "
418 						"vector %d", vector);
419 			break;
420 		}
421 
422 		kvm_make_request(KVM_REQ_EVENT, vcpu);
423 		kvm_vcpu_kick(vcpu);
424 		break;
425 
426 	case APIC_DM_REMRD:
427 		apic_debug("Ignoring delivery mode 3\n");
428 		break;
429 
430 	case APIC_DM_SMI:
431 		apic_debug("Ignoring guest SMI\n");
432 		break;
433 
434 	case APIC_DM_NMI:
435 		result = 1;
436 		kvm_inject_nmi(vcpu);
437 		kvm_vcpu_kick(vcpu);
438 		break;
439 
440 	case APIC_DM_INIT:
441 		if (!trig_mode || level) {
442 			result = 1;
443 			vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
444 			kvm_make_request(KVM_REQ_EVENT, vcpu);
445 			kvm_vcpu_kick(vcpu);
446 		} else {
447 			apic_debug("Ignoring de-assert INIT to vcpu %d\n",
448 				   vcpu->vcpu_id);
449 		}
450 		break;
451 
452 	case APIC_DM_STARTUP:
453 		apic_debug("SIPI to vcpu %d vector 0x%02x\n",
454 			   vcpu->vcpu_id, vector);
455 		if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) {
456 			result = 1;
457 			vcpu->arch.sipi_vector = vector;
458 			vcpu->arch.mp_state = KVM_MP_STATE_SIPI_RECEIVED;
459 			kvm_make_request(KVM_REQ_EVENT, vcpu);
460 			kvm_vcpu_kick(vcpu);
461 		}
462 		break;
463 
464 	case APIC_DM_EXTINT:
465 		/*
466 		 * Should only be called by kvm_apic_local_deliver() with LVT0,
467 		 * before NMI watchdog was enabled. Already handled by
468 		 * kvm_apic_accept_pic_intr().
469 		 */
470 		break;
471 
472 	default:
473 		printk(KERN_ERR "TODO: unsupported delivery mode %x\n",
474 		       delivery_mode);
475 		break;
476 	}
477 	return result;
478 }
479 
480 int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2)
481 {
482 	return vcpu1->arch.apic_arb_prio - vcpu2->arch.apic_arb_prio;
483 }
484 
485 static void apic_set_eoi(struct kvm_lapic *apic)
486 {
487 	int vector = apic_find_highest_isr(apic);
488 	/*
489 	 * Not every write EOI will has corresponding ISR,
490 	 * one example is when Kernel check timer on setup_IO_APIC
491 	 */
492 	if (vector == -1)
493 		return;
494 
495 	apic_clear_vector(vector, apic->regs + APIC_ISR);
496 	apic_update_ppr(apic);
497 
498 	if (!(apic_get_reg(apic, APIC_SPIV) & APIC_SPIV_DIRECTED_EOI) &&
499 	    kvm_ioapic_handles_vector(apic->vcpu->kvm, vector)) {
500 		int trigger_mode;
501 		if (apic_test_vector(vector, apic->regs + APIC_TMR))
502 			trigger_mode = IOAPIC_LEVEL_TRIG;
503 		else
504 			trigger_mode = IOAPIC_EDGE_TRIG;
505 		kvm_ioapic_update_eoi(apic->vcpu->kvm, vector, trigger_mode);
506 	}
507 	kvm_make_request(KVM_REQ_EVENT, apic->vcpu);
508 }
509 
510 static void apic_send_ipi(struct kvm_lapic *apic)
511 {
512 	u32 icr_low = apic_get_reg(apic, APIC_ICR);
513 	u32 icr_high = apic_get_reg(apic, APIC_ICR2);
514 	struct kvm_lapic_irq irq;
515 
516 	irq.vector = icr_low & APIC_VECTOR_MASK;
517 	irq.delivery_mode = icr_low & APIC_MODE_MASK;
518 	irq.dest_mode = icr_low & APIC_DEST_MASK;
519 	irq.level = icr_low & APIC_INT_ASSERT;
520 	irq.trig_mode = icr_low & APIC_INT_LEVELTRIG;
521 	irq.shorthand = icr_low & APIC_SHORT_MASK;
522 	if (apic_x2apic_mode(apic))
523 		irq.dest_id = icr_high;
524 	else
525 		irq.dest_id = GET_APIC_DEST_FIELD(icr_high);
526 
527 	trace_kvm_apic_ipi(icr_low, irq.dest_id);
528 
529 	apic_debug("icr_high 0x%x, icr_low 0x%x, "
530 		   "short_hand 0x%x, dest 0x%x, trig_mode 0x%x, level 0x%x, "
531 		   "dest_mode 0x%x, delivery_mode 0x%x, vector 0x%x\n",
532 		   icr_high, icr_low, irq.shorthand, irq.dest_id,
533 		   irq.trig_mode, irq.level, irq.dest_mode, irq.delivery_mode,
534 		   irq.vector);
535 
536 	kvm_irq_delivery_to_apic(apic->vcpu->kvm, apic, &irq);
537 }
538 
539 static u32 apic_get_tmcct(struct kvm_lapic *apic)
540 {
541 	ktime_t remaining;
542 	s64 ns;
543 	u32 tmcct;
544 
545 	ASSERT(apic != NULL);
546 
547 	/* if initial count is 0, current count should also be 0 */
548 	if (apic_get_reg(apic, APIC_TMICT) == 0)
549 		return 0;
550 
551 	remaining = hrtimer_get_remaining(&apic->lapic_timer.timer);
552 	if (ktime_to_ns(remaining) < 0)
553 		remaining = ktime_set(0, 0);
554 
555 	ns = mod_64(ktime_to_ns(remaining), apic->lapic_timer.period);
556 	tmcct = div64_u64(ns,
557 			 (APIC_BUS_CYCLE_NS * apic->divide_count));
558 
559 	return tmcct;
560 }
561 
562 static void __report_tpr_access(struct kvm_lapic *apic, bool write)
563 {
564 	struct kvm_vcpu *vcpu = apic->vcpu;
565 	struct kvm_run *run = vcpu->run;
566 
567 	kvm_make_request(KVM_REQ_REPORT_TPR_ACCESS, vcpu);
568 	run->tpr_access.rip = kvm_rip_read(vcpu);
569 	run->tpr_access.is_write = write;
570 }
571 
572 static inline void report_tpr_access(struct kvm_lapic *apic, bool write)
573 {
574 	if (apic->vcpu->arch.tpr_access_reporting)
575 		__report_tpr_access(apic, write);
576 }
577 
578 static u32 __apic_read(struct kvm_lapic *apic, unsigned int offset)
579 {
580 	u32 val = 0;
581 
582 	if (offset >= LAPIC_MMIO_LENGTH)
583 		return 0;
584 
585 	switch (offset) {
586 	case APIC_ID:
587 		if (apic_x2apic_mode(apic))
588 			val = kvm_apic_id(apic);
589 		else
590 			val = kvm_apic_id(apic) << 24;
591 		break;
592 	case APIC_ARBPRI:
593 		apic_debug("Access APIC ARBPRI register which is for P6\n");
594 		break;
595 
596 	case APIC_TMCCT:	/* Timer CCR */
597 		if (apic_lvtt_tscdeadline(apic))
598 			return 0;
599 
600 		val = apic_get_tmcct(apic);
601 		break;
602 
603 	case APIC_TASKPRI:
604 		report_tpr_access(apic, false);
605 		/* fall thru */
606 	default:
607 		apic_update_ppr(apic);
608 		val = apic_get_reg(apic, offset);
609 		break;
610 	}
611 
612 	return val;
613 }
614 
615 static inline struct kvm_lapic *to_lapic(struct kvm_io_device *dev)
616 {
617 	return container_of(dev, struct kvm_lapic, dev);
618 }
619 
620 static int apic_reg_read(struct kvm_lapic *apic, u32 offset, int len,
621 		void *data)
622 {
623 	unsigned char alignment = offset & 0xf;
624 	u32 result;
625 	/* this bitmask has a bit cleared for each reserver register */
626 	static const u64 rmask = 0x43ff01ffffffe70cULL;
627 
628 	if ((alignment + len) > 4) {
629 		apic_debug("KVM_APIC_READ: alignment error %x %d\n",
630 			   offset, len);
631 		return 1;
632 	}
633 
634 	if (offset > 0x3f0 || !(rmask & (1ULL << (offset >> 4)))) {
635 		apic_debug("KVM_APIC_READ: read reserved register %x\n",
636 			   offset);
637 		return 1;
638 	}
639 
640 	result = __apic_read(apic, offset & ~0xf);
641 
642 	trace_kvm_apic_read(offset, result);
643 
644 	switch (len) {
645 	case 1:
646 	case 2:
647 	case 4:
648 		memcpy(data, (char *)&result + alignment, len);
649 		break;
650 	default:
651 		printk(KERN_ERR "Local APIC read with len = %x, "
652 		       "should be 1,2, or 4 instead\n", len);
653 		break;
654 	}
655 	return 0;
656 }
657 
658 static int apic_mmio_in_range(struct kvm_lapic *apic, gpa_t addr)
659 {
660 	return apic_hw_enabled(apic) &&
661 	    addr >= apic->base_address &&
662 	    addr < apic->base_address + LAPIC_MMIO_LENGTH;
663 }
664 
665 static int apic_mmio_read(struct kvm_io_device *this,
666 			   gpa_t address, int len, void *data)
667 {
668 	struct kvm_lapic *apic = to_lapic(this);
669 	u32 offset = address - apic->base_address;
670 
671 	if (!apic_mmio_in_range(apic, address))
672 		return -EOPNOTSUPP;
673 
674 	apic_reg_read(apic, offset, len, data);
675 
676 	return 0;
677 }
678 
679 static void update_divide_count(struct kvm_lapic *apic)
680 {
681 	u32 tmp1, tmp2, tdcr;
682 
683 	tdcr = apic_get_reg(apic, APIC_TDCR);
684 	tmp1 = tdcr & 0xf;
685 	tmp2 = ((tmp1 & 0x3) | ((tmp1 & 0x8) >> 1)) + 1;
686 	apic->divide_count = 0x1 << (tmp2 & 0x7);
687 
688 	apic_debug("timer divide count is 0x%x\n",
689 				   apic->divide_count);
690 }
691 
692 static void start_apic_timer(struct kvm_lapic *apic)
693 {
694 	ktime_t now;
695 	atomic_set(&apic->lapic_timer.pending, 0);
696 
697 	if (apic_lvtt_period(apic) || apic_lvtt_oneshot(apic)) {
698 		/* lapic timer in oneshot or peroidic mode */
699 		now = apic->lapic_timer.timer.base->get_time();
700 		apic->lapic_timer.period = (u64)apic_get_reg(apic, APIC_TMICT)
701 			    * APIC_BUS_CYCLE_NS * apic->divide_count;
702 
703 		if (!apic->lapic_timer.period)
704 			return;
705 		/*
706 		 * Do not allow the guest to program periodic timers with small
707 		 * interval, since the hrtimers are not throttled by the host
708 		 * scheduler.
709 		 */
710 		if (apic_lvtt_period(apic)) {
711 			s64 min_period = min_timer_period_us * 1000LL;
712 
713 			if (apic->lapic_timer.period < min_period) {
714 				pr_info_ratelimited(
715 				    "kvm: vcpu %i: requested %lld ns "
716 				    "lapic timer period limited to %lld ns\n",
717 				    apic->vcpu->vcpu_id,
718 				    apic->lapic_timer.period, min_period);
719 				apic->lapic_timer.period = min_period;
720 			}
721 		}
722 
723 		hrtimer_start(&apic->lapic_timer.timer,
724 			      ktime_add_ns(now, apic->lapic_timer.period),
725 			      HRTIMER_MODE_ABS);
726 
727 		apic_debug("%s: bus cycle is %" PRId64 "ns, now 0x%016"
728 			   PRIx64 ", "
729 			   "timer initial count 0x%x, period %lldns, "
730 			   "expire @ 0x%016" PRIx64 ".\n", __func__,
731 			   APIC_BUS_CYCLE_NS, ktime_to_ns(now),
732 			   apic_get_reg(apic, APIC_TMICT),
733 			   apic->lapic_timer.period,
734 			   ktime_to_ns(ktime_add_ns(now,
735 					apic->lapic_timer.period)));
736 	} else if (apic_lvtt_tscdeadline(apic)) {
737 		/* lapic timer in tsc deadline mode */
738 		u64 guest_tsc, tscdeadline = apic->lapic_timer.tscdeadline;
739 		u64 ns = 0;
740 		struct kvm_vcpu *vcpu = apic->vcpu;
741 		unsigned long this_tsc_khz = vcpu->arch.virtual_tsc_khz;
742 		unsigned long flags;
743 
744 		if (unlikely(!tscdeadline || !this_tsc_khz))
745 			return;
746 
747 		local_irq_save(flags);
748 
749 		now = apic->lapic_timer.timer.base->get_time();
750 		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu);
751 		if (likely(tscdeadline > guest_tsc)) {
752 			ns = (tscdeadline - guest_tsc) * 1000000ULL;
753 			do_div(ns, this_tsc_khz);
754 		}
755 		hrtimer_start(&apic->lapic_timer.timer,
756 			ktime_add_ns(now, ns), HRTIMER_MODE_ABS);
757 
758 		local_irq_restore(flags);
759 	}
760 }
761 
762 static void apic_manage_nmi_watchdog(struct kvm_lapic *apic, u32 lvt0_val)
763 {
764 	int nmi_wd_enabled = apic_lvt_nmi_mode(apic_get_reg(apic, APIC_LVT0));
765 
766 	if (apic_lvt_nmi_mode(lvt0_val)) {
767 		if (!nmi_wd_enabled) {
768 			apic_debug("Receive NMI setting on APIC_LVT0 "
769 				   "for cpu %d\n", apic->vcpu->vcpu_id);
770 			apic->vcpu->kvm->arch.vapics_in_nmi_mode++;
771 		}
772 	} else if (nmi_wd_enabled)
773 		apic->vcpu->kvm->arch.vapics_in_nmi_mode--;
774 }
775 
776 static int apic_reg_write(struct kvm_lapic *apic, u32 reg, u32 val)
777 {
778 	int ret = 0;
779 
780 	trace_kvm_apic_write(reg, val);
781 
782 	switch (reg) {
783 	case APIC_ID:		/* Local APIC ID */
784 		if (!apic_x2apic_mode(apic))
785 			apic_set_reg(apic, APIC_ID, val);
786 		else
787 			ret = 1;
788 		break;
789 
790 	case APIC_TASKPRI:
791 		report_tpr_access(apic, true);
792 		apic_set_tpr(apic, val & 0xff);
793 		break;
794 
795 	case APIC_EOI:
796 		apic_set_eoi(apic);
797 		break;
798 
799 	case APIC_LDR:
800 		if (!apic_x2apic_mode(apic))
801 			apic_set_reg(apic, APIC_LDR, val & APIC_LDR_MASK);
802 		else
803 			ret = 1;
804 		break;
805 
806 	case APIC_DFR:
807 		if (!apic_x2apic_mode(apic))
808 			apic_set_reg(apic, APIC_DFR, val | 0x0FFFFFFF);
809 		else
810 			ret = 1;
811 		break;
812 
813 	case APIC_SPIV: {
814 		u32 mask = 0x3ff;
815 		if (apic_get_reg(apic, APIC_LVR) & APIC_LVR_DIRECTED_EOI)
816 			mask |= APIC_SPIV_DIRECTED_EOI;
817 		apic_set_reg(apic, APIC_SPIV, val & mask);
818 		if (!(val & APIC_SPIV_APIC_ENABLED)) {
819 			int i;
820 			u32 lvt_val;
821 
822 			for (i = 0; i < APIC_LVT_NUM; i++) {
823 				lvt_val = apic_get_reg(apic,
824 						       APIC_LVTT + 0x10 * i);
825 				apic_set_reg(apic, APIC_LVTT + 0x10 * i,
826 					     lvt_val | APIC_LVT_MASKED);
827 			}
828 			atomic_set(&apic->lapic_timer.pending, 0);
829 
830 		}
831 		break;
832 	}
833 	case APIC_ICR:
834 		/* No delay here, so we always clear the pending bit */
835 		apic_set_reg(apic, APIC_ICR, val & ~(1 << 12));
836 		apic_send_ipi(apic);
837 		break;
838 
839 	case APIC_ICR2:
840 		if (!apic_x2apic_mode(apic))
841 			val &= 0xff000000;
842 		apic_set_reg(apic, APIC_ICR2, val);
843 		break;
844 
845 	case APIC_LVT0:
846 		apic_manage_nmi_watchdog(apic, val);
847 	case APIC_LVTTHMR:
848 	case APIC_LVTPC:
849 	case APIC_LVT1:
850 	case APIC_LVTERR:
851 		/* TODO: Check vector */
852 		if (!apic_sw_enabled(apic))
853 			val |= APIC_LVT_MASKED;
854 
855 		val &= apic_lvt_mask[(reg - APIC_LVTT) >> 4];
856 		apic_set_reg(apic, reg, val);
857 
858 		break;
859 
860 	case APIC_LVTT:
861 		if ((apic_get_reg(apic, APIC_LVTT) &
862 		    apic->lapic_timer.timer_mode_mask) !=
863 		   (val & apic->lapic_timer.timer_mode_mask))
864 			hrtimer_cancel(&apic->lapic_timer.timer);
865 
866 		if (!apic_sw_enabled(apic))
867 			val |= APIC_LVT_MASKED;
868 		val &= (apic_lvt_mask[0] | apic->lapic_timer.timer_mode_mask);
869 		apic_set_reg(apic, APIC_LVTT, val);
870 		break;
871 
872 	case APIC_TMICT:
873 		if (apic_lvtt_tscdeadline(apic))
874 			break;
875 
876 		hrtimer_cancel(&apic->lapic_timer.timer);
877 		apic_set_reg(apic, APIC_TMICT, val);
878 		start_apic_timer(apic);
879 		break;
880 
881 	case APIC_TDCR:
882 		if (val & 4)
883 			apic_debug("KVM_WRITE:TDCR %x\n", val);
884 		apic_set_reg(apic, APIC_TDCR, val);
885 		update_divide_count(apic);
886 		break;
887 
888 	case APIC_ESR:
889 		if (apic_x2apic_mode(apic) && val != 0) {
890 			apic_debug("KVM_WRITE:ESR not zero %x\n", val);
891 			ret = 1;
892 		}
893 		break;
894 
895 	case APIC_SELF_IPI:
896 		if (apic_x2apic_mode(apic)) {
897 			apic_reg_write(apic, APIC_ICR, 0x40000 | (val & 0xff));
898 		} else
899 			ret = 1;
900 		break;
901 	default:
902 		ret = 1;
903 		break;
904 	}
905 	if (ret)
906 		apic_debug("Local APIC Write to read-only register %x\n", reg);
907 	return ret;
908 }
909 
910 static int apic_mmio_write(struct kvm_io_device *this,
911 			    gpa_t address, int len, const void *data)
912 {
913 	struct kvm_lapic *apic = to_lapic(this);
914 	unsigned int offset = address - apic->base_address;
915 	u32 val;
916 
917 	if (!apic_mmio_in_range(apic, address))
918 		return -EOPNOTSUPP;
919 
920 	/*
921 	 * APIC register must be aligned on 128-bits boundary.
922 	 * 32/64/128 bits registers must be accessed thru 32 bits.
923 	 * Refer SDM 8.4.1
924 	 */
925 	if (len != 4 || (offset & 0xf)) {
926 		/* Don't shout loud, $infamous_os would cause only noise. */
927 		apic_debug("apic write: bad size=%d %lx\n", len, (long)address);
928 		return 0;
929 	}
930 
931 	val = *(u32*)data;
932 
933 	/* too common printing */
934 	if (offset != APIC_EOI)
935 		apic_debug("%s: offset 0x%x with length 0x%x, and value is "
936 			   "0x%x\n", __func__, offset, len, val);
937 
938 	apic_reg_write(apic, offset & 0xff0, val);
939 
940 	return 0;
941 }
942 
943 void kvm_lapic_set_eoi(struct kvm_vcpu *vcpu)
944 {
945 	struct kvm_lapic *apic = vcpu->arch.apic;
946 
947 	if (apic)
948 		apic_reg_write(vcpu->arch.apic, APIC_EOI, 0);
949 }
950 EXPORT_SYMBOL_GPL(kvm_lapic_set_eoi);
951 
952 void kvm_free_lapic(struct kvm_vcpu *vcpu)
953 {
954 	if (!vcpu->arch.apic)
955 		return;
956 
957 	hrtimer_cancel(&vcpu->arch.apic->lapic_timer.timer);
958 
959 	if (vcpu->arch.apic->regs)
960 		free_page((unsigned long)vcpu->arch.apic->regs);
961 
962 	kfree(vcpu->arch.apic);
963 }
964 
965 /*
966  *----------------------------------------------------------------------
967  * LAPIC interface
968  *----------------------------------------------------------------------
969  */
970 
971 u64 kvm_get_lapic_tscdeadline_msr(struct kvm_vcpu *vcpu)
972 {
973 	struct kvm_lapic *apic = vcpu->arch.apic;
974 	if (!apic)
975 		return 0;
976 
977 	if (apic_lvtt_oneshot(apic) || apic_lvtt_period(apic))
978 		return 0;
979 
980 	return apic->lapic_timer.tscdeadline;
981 }
982 
983 void kvm_set_lapic_tscdeadline_msr(struct kvm_vcpu *vcpu, u64 data)
984 {
985 	struct kvm_lapic *apic = vcpu->arch.apic;
986 	if (!apic)
987 		return;
988 
989 	if (apic_lvtt_oneshot(apic) || apic_lvtt_period(apic))
990 		return;
991 
992 	hrtimer_cancel(&apic->lapic_timer.timer);
993 	apic->lapic_timer.tscdeadline = data;
994 	start_apic_timer(apic);
995 }
996 
997 void kvm_lapic_set_tpr(struct kvm_vcpu *vcpu, unsigned long cr8)
998 {
999 	struct kvm_lapic *apic = vcpu->arch.apic;
1000 
1001 	if (!apic)
1002 		return;
1003 	apic_set_tpr(apic, ((cr8 & 0x0f) << 4)
1004 		     | (apic_get_reg(apic, APIC_TASKPRI) & 4));
1005 }
1006 
1007 u64 kvm_lapic_get_cr8(struct kvm_vcpu *vcpu)
1008 {
1009 	struct kvm_lapic *apic = vcpu->arch.apic;
1010 	u64 tpr;
1011 
1012 	if (!apic)
1013 		return 0;
1014 	tpr = (u64) apic_get_reg(apic, APIC_TASKPRI);
1015 
1016 	return (tpr & 0xf0) >> 4;
1017 }
1018 
1019 void kvm_lapic_set_base(struct kvm_vcpu *vcpu, u64 value)
1020 {
1021 	struct kvm_lapic *apic = vcpu->arch.apic;
1022 
1023 	if (!apic) {
1024 		value |= MSR_IA32_APICBASE_BSP;
1025 		vcpu->arch.apic_base = value;
1026 		return;
1027 	}
1028 
1029 	if (!kvm_vcpu_is_bsp(apic->vcpu))
1030 		value &= ~MSR_IA32_APICBASE_BSP;
1031 
1032 	vcpu->arch.apic_base = value;
1033 	if (apic_x2apic_mode(apic)) {
1034 		u32 id = kvm_apic_id(apic);
1035 		u32 ldr = ((id & ~0xf) << 16) | (1 << (id & 0xf));
1036 		apic_set_reg(apic, APIC_LDR, ldr);
1037 	}
1038 	apic->base_address = apic->vcpu->arch.apic_base &
1039 			     MSR_IA32_APICBASE_BASE;
1040 
1041 	/* with FSB delivery interrupt, we can restart APIC functionality */
1042 	apic_debug("apic base msr is 0x%016" PRIx64 ", and base address is "
1043 		   "0x%lx.\n", apic->vcpu->arch.apic_base, apic->base_address);
1044 
1045 }
1046 
1047 void kvm_lapic_reset(struct kvm_vcpu *vcpu)
1048 {
1049 	struct kvm_lapic *apic;
1050 	int i;
1051 
1052 	apic_debug("%s\n", __func__);
1053 
1054 	ASSERT(vcpu);
1055 	apic = vcpu->arch.apic;
1056 	ASSERT(apic != NULL);
1057 
1058 	/* Stop the timer in case it's a reset to an active apic */
1059 	hrtimer_cancel(&apic->lapic_timer.timer);
1060 
1061 	apic_set_reg(apic, APIC_ID, vcpu->vcpu_id << 24);
1062 	kvm_apic_set_version(apic->vcpu);
1063 
1064 	for (i = 0; i < APIC_LVT_NUM; i++)
1065 		apic_set_reg(apic, APIC_LVTT + 0x10 * i, APIC_LVT_MASKED);
1066 	apic_set_reg(apic, APIC_LVT0,
1067 		     SET_APIC_DELIVERY_MODE(0, APIC_MODE_EXTINT));
1068 
1069 	apic_set_reg(apic, APIC_DFR, 0xffffffffU);
1070 	apic_set_reg(apic, APIC_SPIV, 0xff);
1071 	apic_set_reg(apic, APIC_TASKPRI, 0);
1072 	apic_set_reg(apic, APIC_LDR, 0);
1073 	apic_set_reg(apic, APIC_ESR, 0);
1074 	apic_set_reg(apic, APIC_ICR, 0);
1075 	apic_set_reg(apic, APIC_ICR2, 0);
1076 	apic_set_reg(apic, APIC_TDCR, 0);
1077 	apic_set_reg(apic, APIC_TMICT, 0);
1078 	for (i = 0; i < 8; i++) {
1079 		apic_set_reg(apic, APIC_IRR + 0x10 * i, 0);
1080 		apic_set_reg(apic, APIC_ISR + 0x10 * i, 0);
1081 		apic_set_reg(apic, APIC_TMR + 0x10 * i, 0);
1082 	}
1083 	apic->irr_pending = false;
1084 	update_divide_count(apic);
1085 	atomic_set(&apic->lapic_timer.pending, 0);
1086 	if (kvm_vcpu_is_bsp(vcpu))
1087 		vcpu->arch.apic_base |= MSR_IA32_APICBASE_BSP;
1088 	apic_update_ppr(apic);
1089 
1090 	vcpu->arch.apic_arb_prio = 0;
1091 	vcpu->arch.apic_attention = 0;
1092 
1093 	apic_debug(KERN_INFO "%s: vcpu=%p, id=%d, base_msr="
1094 		   "0x%016" PRIx64 ", base_address=0x%0lx.\n", __func__,
1095 		   vcpu, kvm_apic_id(apic),
1096 		   vcpu->arch.apic_base, apic->base_address);
1097 }
1098 
1099 bool kvm_apic_present(struct kvm_vcpu *vcpu)
1100 {
1101 	return vcpu->arch.apic && apic_hw_enabled(vcpu->arch.apic);
1102 }
1103 
1104 int kvm_lapic_enabled(struct kvm_vcpu *vcpu)
1105 {
1106 	return kvm_apic_present(vcpu) && apic_sw_enabled(vcpu->arch.apic);
1107 }
1108 
1109 /*
1110  *----------------------------------------------------------------------
1111  * timer interface
1112  *----------------------------------------------------------------------
1113  */
1114 
1115 static bool lapic_is_periodic(struct kvm_timer *ktimer)
1116 {
1117 	struct kvm_lapic *apic = container_of(ktimer, struct kvm_lapic,
1118 					      lapic_timer);
1119 	return apic_lvtt_period(apic);
1120 }
1121 
1122 int apic_has_pending_timer(struct kvm_vcpu *vcpu)
1123 {
1124 	struct kvm_lapic *lapic = vcpu->arch.apic;
1125 
1126 	if (lapic && apic_enabled(lapic) && apic_lvt_enabled(lapic, APIC_LVTT))
1127 		return atomic_read(&lapic->lapic_timer.pending);
1128 
1129 	return 0;
1130 }
1131 
1132 int kvm_apic_local_deliver(struct kvm_lapic *apic, int lvt_type)
1133 {
1134 	u32 reg = apic_get_reg(apic, lvt_type);
1135 	int vector, mode, trig_mode;
1136 
1137 	if (apic_hw_enabled(apic) && !(reg & APIC_LVT_MASKED)) {
1138 		vector = reg & APIC_VECTOR_MASK;
1139 		mode = reg & APIC_MODE_MASK;
1140 		trig_mode = reg & APIC_LVT_LEVEL_TRIGGER;
1141 		return __apic_accept_irq(apic, mode, vector, 1, trig_mode);
1142 	}
1143 	return 0;
1144 }
1145 
1146 void kvm_apic_nmi_wd_deliver(struct kvm_vcpu *vcpu)
1147 {
1148 	struct kvm_lapic *apic = vcpu->arch.apic;
1149 
1150 	if (apic)
1151 		kvm_apic_local_deliver(apic, APIC_LVT0);
1152 }
1153 
1154 static struct kvm_timer_ops lapic_timer_ops = {
1155 	.is_periodic = lapic_is_periodic,
1156 };
1157 
1158 static const struct kvm_io_device_ops apic_mmio_ops = {
1159 	.read     = apic_mmio_read,
1160 	.write    = apic_mmio_write,
1161 };
1162 
1163 int kvm_create_lapic(struct kvm_vcpu *vcpu)
1164 {
1165 	struct kvm_lapic *apic;
1166 
1167 	ASSERT(vcpu != NULL);
1168 	apic_debug("apic_init %d\n", vcpu->vcpu_id);
1169 
1170 	apic = kzalloc(sizeof(*apic), GFP_KERNEL);
1171 	if (!apic)
1172 		goto nomem;
1173 
1174 	vcpu->arch.apic = apic;
1175 
1176 	apic->regs = (void *)get_zeroed_page(GFP_KERNEL);
1177 	if (!apic->regs) {
1178 		printk(KERN_ERR "malloc apic regs error for vcpu %x\n",
1179 		       vcpu->vcpu_id);
1180 		goto nomem_free_apic;
1181 	}
1182 	apic->vcpu = vcpu;
1183 
1184 	hrtimer_init(&apic->lapic_timer.timer, CLOCK_MONOTONIC,
1185 		     HRTIMER_MODE_ABS);
1186 	apic->lapic_timer.timer.function = kvm_timer_fn;
1187 	apic->lapic_timer.t_ops = &lapic_timer_ops;
1188 	apic->lapic_timer.kvm = vcpu->kvm;
1189 	apic->lapic_timer.vcpu = vcpu;
1190 
1191 	apic->base_address = APIC_DEFAULT_PHYS_BASE;
1192 	vcpu->arch.apic_base = APIC_DEFAULT_PHYS_BASE;
1193 
1194 	kvm_lapic_reset(vcpu);
1195 	kvm_iodevice_init(&apic->dev, &apic_mmio_ops);
1196 
1197 	return 0;
1198 nomem_free_apic:
1199 	kfree(apic);
1200 nomem:
1201 	return -ENOMEM;
1202 }
1203 
1204 int kvm_apic_has_interrupt(struct kvm_vcpu *vcpu)
1205 {
1206 	struct kvm_lapic *apic = vcpu->arch.apic;
1207 	int highest_irr;
1208 
1209 	if (!apic || !apic_enabled(apic))
1210 		return -1;
1211 
1212 	apic_update_ppr(apic);
1213 	highest_irr = apic_find_highest_irr(apic);
1214 	if ((highest_irr == -1) ||
1215 	    ((highest_irr & 0xF0) <= apic_get_reg(apic, APIC_PROCPRI)))
1216 		return -1;
1217 	return highest_irr;
1218 }
1219 
1220 int kvm_apic_accept_pic_intr(struct kvm_vcpu *vcpu)
1221 {
1222 	u32 lvt0 = apic_get_reg(vcpu->arch.apic, APIC_LVT0);
1223 	int r = 0;
1224 
1225 	if (!apic_hw_enabled(vcpu->arch.apic))
1226 		r = 1;
1227 	if ((lvt0 & APIC_LVT_MASKED) == 0 &&
1228 	    GET_APIC_DELIVERY_MODE(lvt0) == APIC_MODE_EXTINT)
1229 		r = 1;
1230 	return r;
1231 }
1232 
1233 void kvm_inject_apic_timer_irqs(struct kvm_vcpu *vcpu)
1234 {
1235 	struct kvm_lapic *apic = vcpu->arch.apic;
1236 
1237 	if (apic && atomic_read(&apic->lapic_timer.pending) > 0) {
1238 		if (kvm_apic_local_deliver(apic, APIC_LVTT))
1239 			atomic_dec(&apic->lapic_timer.pending);
1240 	}
1241 }
1242 
1243 int kvm_get_apic_interrupt(struct kvm_vcpu *vcpu)
1244 {
1245 	int vector = kvm_apic_has_interrupt(vcpu);
1246 	struct kvm_lapic *apic = vcpu->arch.apic;
1247 
1248 	if (vector == -1)
1249 		return -1;
1250 
1251 	apic_set_vector(vector, apic->regs + APIC_ISR);
1252 	apic_update_ppr(apic);
1253 	apic_clear_irr(vector, apic);
1254 	return vector;
1255 }
1256 
1257 void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu)
1258 {
1259 	struct kvm_lapic *apic = vcpu->arch.apic;
1260 
1261 	apic->base_address = vcpu->arch.apic_base &
1262 			     MSR_IA32_APICBASE_BASE;
1263 	kvm_apic_set_version(vcpu);
1264 
1265 	apic_update_ppr(apic);
1266 	hrtimer_cancel(&apic->lapic_timer.timer);
1267 	update_divide_count(apic);
1268 	start_apic_timer(apic);
1269 	apic->irr_pending = true;
1270 	kvm_make_request(KVM_REQ_EVENT, vcpu);
1271 }
1272 
1273 void __kvm_migrate_apic_timer(struct kvm_vcpu *vcpu)
1274 {
1275 	struct kvm_lapic *apic = vcpu->arch.apic;
1276 	struct hrtimer *timer;
1277 
1278 	if (!apic)
1279 		return;
1280 
1281 	timer = &apic->lapic_timer.timer;
1282 	if (hrtimer_cancel(timer))
1283 		hrtimer_start_expires(timer, HRTIMER_MODE_ABS);
1284 }
1285 
1286 void kvm_lapic_sync_from_vapic(struct kvm_vcpu *vcpu)
1287 {
1288 	u32 data;
1289 	void *vapic;
1290 
1291 	if (!test_bit(KVM_APIC_CHECK_VAPIC, &vcpu->arch.apic_attention))
1292 		return;
1293 
1294 	vapic = kmap_atomic(vcpu->arch.apic->vapic_page);
1295 	data = *(u32 *)(vapic + offset_in_page(vcpu->arch.apic->vapic_addr));
1296 	kunmap_atomic(vapic);
1297 
1298 	apic_set_tpr(vcpu->arch.apic, data & 0xff);
1299 }
1300 
1301 void kvm_lapic_sync_to_vapic(struct kvm_vcpu *vcpu)
1302 {
1303 	u32 data, tpr;
1304 	int max_irr, max_isr;
1305 	struct kvm_lapic *apic;
1306 	void *vapic;
1307 
1308 	if (!test_bit(KVM_APIC_CHECK_VAPIC, &vcpu->arch.apic_attention))
1309 		return;
1310 
1311 	apic = vcpu->arch.apic;
1312 	tpr = apic_get_reg(apic, APIC_TASKPRI) & 0xff;
1313 	max_irr = apic_find_highest_irr(apic);
1314 	if (max_irr < 0)
1315 		max_irr = 0;
1316 	max_isr = apic_find_highest_isr(apic);
1317 	if (max_isr < 0)
1318 		max_isr = 0;
1319 	data = (tpr & 0xff) | ((max_isr & 0xf0) << 8) | (max_irr << 24);
1320 
1321 	vapic = kmap_atomic(vcpu->arch.apic->vapic_page);
1322 	*(u32 *)(vapic + offset_in_page(vcpu->arch.apic->vapic_addr)) = data;
1323 	kunmap_atomic(vapic);
1324 }
1325 
1326 void kvm_lapic_set_vapic_addr(struct kvm_vcpu *vcpu, gpa_t vapic_addr)
1327 {
1328 	vcpu->arch.apic->vapic_addr = vapic_addr;
1329 	if (vapic_addr)
1330 		__set_bit(KVM_APIC_CHECK_VAPIC, &vcpu->arch.apic_attention);
1331 	else
1332 		__clear_bit(KVM_APIC_CHECK_VAPIC, &vcpu->arch.apic_attention);
1333 }
1334 
1335 int kvm_x2apic_msr_write(struct kvm_vcpu *vcpu, u32 msr, u64 data)
1336 {
1337 	struct kvm_lapic *apic = vcpu->arch.apic;
1338 	u32 reg = (msr - APIC_BASE_MSR) << 4;
1339 
1340 	if (!irqchip_in_kernel(vcpu->kvm) || !apic_x2apic_mode(apic))
1341 		return 1;
1342 
1343 	/* if this is ICR write vector before command */
1344 	if (msr == 0x830)
1345 		apic_reg_write(apic, APIC_ICR2, (u32)(data >> 32));
1346 	return apic_reg_write(apic, reg, (u32)data);
1347 }
1348 
1349 int kvm_x2apic_msr_read(struct kvm_vcpu *vcpu, u32 msr, u64 *data)
1350 {
1351 	struct kvm_lapic *apic = vcpu->arch.apic;
1352 	u32 reg = (msr - APIC_BASE_MSR) << 4, low, high = 0;
1353 
1354 	if (!irqchip_in_kernel(vcpu->kvm) || !apic_x2apic_mode(apic))
1355 		return 1;
1356 
1357 	if (apic_reg_read(apic, reg, 4, &low))
1358 		return 1;
1359 	if (msr == 0x830)
1360 		apic_reg_read(apic, APIC_ICR2, 4, &high);
1361 
1362 	*data = (((u64)high) << 32) | low;
1363 
1364 	return 0;
1365 }
1366 
1367 int kvm_hv_vapic_msr_write(struct kvm_vcpu *vcpu, u32 reg, u64 data)
1368 {
1369 	struct kvm_lapic *apic = vcpu->arch.apic;
1370 
1371 	if (!irqchip_in_kernel(vcpu->kvm))
1372 		return 1;
1373 
1374 	/* if this is ICR write vector before command */
1375 	if (reg == APIC_ICR)
1376 		apic_reg_write(apic, APIC_ICR2, (u32)(data >> 32));
1377 	return apic_reg_write(apic, reg, (u32)data);
1378 }
1379 
1380 int kvm_hv_vapic_msr_read(struct kvm_vcpu *vcpu, u32 reg, u64 *data)
1381 {
1382 	struct kvm_lapic *apic = vcpu->arch.apic;
1383 	u32 low, high = 0;
1384 
1385 	if (!irqchip_in_kernel(vcpu->kvm))
1386 		return 1;
1387 
1388 	if (apic_reg_read(apic, reg, 4, &low))
1389 		return 1;
1390 	if (reg == APIC_ICR)
1391 		apic_reg_read(apic, APIC_ICR2, 4, &high);
1392 
1393 	*data = (((u64)high) << 32) | low;
1394 
1395 	return 0;
1396 }
1397