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