]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - arch/x86/kvm/lapic.c
KVM: Update MAINTAINERS for new mailing lists
[mirror_ubuntu-bionic-kernel.git] / arch / x86 / kvm / lapic.c
CommitLineData
97222cc8
ED
1
2/*
3 * Local APIC virtualization
4 *
5 * Copyright (C) 2006 Qumranet, Inc.
6 * Copyright (C) 2007 Novell
7 * Copyright (C) 2007 Intel
8 *
9 * Authors:
10 * Dor Laor <dor.laor@qumranet.com>
11 * Gregory Haskins <ghaskins@novell.com>
12 * Yaozu (Eddie) Dong <eddie.dong@intel.com>
13 *
14 * Based on Xen 3.1 code, Copyright (c) 2004, Intel Corporation.
15 *
16 * This work is licensed under the terms of the GNU GPL, version 2. See
17 * the COPYING file in the top-level directory.
18 */
19
edf88417 20#include <linux/kvm_host.h>
97222cc8
ED
21#include <linux/kvm.h>
22#include <linux/mm.h>
23#include <linux/highmem.h>
24#include <linux/smp.h>
25#include <linux/hrtimer.h>
26#include <linux/io.h>
27#include <linux/module.h>
6f6d6a1a 28#include <linux/math64.h>
97222cc8
ED
29#include <asm/processor.h>
30#include <asm/msr.h>
31#include <asm/page.h>
32#include <asm/current.h>
33#include <asm/apicdef.h>
34#include <asm/atomic.h>
97222cc8
ED
35#include "irq.h"
36
37#define PRId64 "d"
38#define PRIx64 "llx"
39#define PRIu64 "u"
40#define PRIo64 "o"
41
42#define APIC_BUS_CYCLE_NS 1
43
44/* #define apic_debug(fmt,arg...) printk(KERN_WARNING fmt,##arg) */
45#define apic_debug(fmt, arg...)
46
47#define APIC_LVT_NUM 6
48/* 14 is the version for Xeon and Pentium 8.4.8*/
49#define APIC_VERSION (0x14UL | ((APIC_LVT_NUM - 1) << 16))
50#define LAPIC_MMIO_LENGTH (1 << 12)
51/* followed define is not in apicdef.h */
52#define APIC_SHORT_MASK 0xc0000
53#define APIC_DEST_NOSHORT 0x0
54#define APIC_DEST_MASK 0x800
55#define MAX_APIC_VECTOR 256
56
57#define VEC_POS(v) ((v) & (32 - 1))
58#define REG_POS(v) (((v) >> 5) << 4)
ad312c7c 59
97222cc8
ED
60static inline u32 apic_get_reg(struct kvm_lapic *apic, int reg_off)
61{
62 return *((u32 *) (apic->regs + reg_off));
63}
64
65static inline void apic_set_reg(struct kvm_lapic *apic, int reg_off, u32 val)
66{
67 *((u32 *) (apic->regs + reg_off)) = val;
68}
69
70static inline int apic_test_and_set_vector(int vec, void *bitmap)
71{
72 return test_and_set_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
73}
74
75static inline int apic_test_and_clear_vector(int vec, void *bitmap)
76{
77 return test_and_clear_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
78}
79
80static inline void apic_set_vector(int vec, void *bitmap)
81{
82 set_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
83}
84
85static inline void apic_clear_vector(int vec, void *bitmap)
86{
87 clear_bit(VEC_POS(vec), (bitmap) + REG_POS(vec));
88}
89
90static inline int apic_hw_enabled(struct kvm_lapic *apic)
91{
ad312c7c 92 return (apic)->vcpu->arch.apic_base & MSR_IA32_APICBASE_ENABLE;
97222cc8
ED
93}
94
95static inline int apic_sw_enabled(struct kvm_lapic *apic)
96{
97 return apic_get_reg(apic, APIC_SPIV) & APIC_SPIV_APIC_ENABLED;
98}
99
100static inline int apic_enabled(struct kvm_lapic *apic)
101{
102 return apic_sw_enabled(apic) && apic_hw_enabled(apic);
103}
104
105#define LVT_MASK \
106 (APIC_LVT_MASKED | APIC_SEND_PENDING | APIC_VECTOR_MASK)
107
108#define LINT_MASK \
109 (LVT_MASK | APIC_MODE_MASK | APIC_INPUT_POLARITY | \
110 APIC_LVT_REMOTE_IRR | APIC_LVT_LEVEL_TRIGGER)
111
112static inline int kvm_apic_id(struct kvm_lapic *apic)
113{
114 return (apic_get_reg(apic, APIC_ID) >> 24) & 0xff;
115}
116
117static inline int apic_lvt_enabled(struct kvm_lapic *apic, int lvt_type)
118{
119 return !(apic_get_reg(apic, lvt_type) & APIC_LVT_MASKED);
120}
121
122static inline int apic_lvt_vector(struct kvm_lapic *apic, int lvt_type)
123{
124 return apic_get_reg(apic, lvt_type) & APIC_VECTOR_MASK;
125}
126
127static inline int apic_lvtt_period(struct kvm_lapic *apic)
128{
129 return apic_get_reg(apic, APIC_LVTT) & APIC_LVT_TIMER_PERIODIC;
130}
131
132static unsigned int apic_lvt_mask[APIC_LVT_NUM] = {
133 LVT_MASK | APIC_LVT_TIMER_PERIODIC, /* LVTT */
134 LVT_MASK | APIC_MODE_MASK, /* LVTTHMR */
135 LVT_MASK | APIC_MODE_MASK, /* LVTPC */
136 LINT_MASK, LINT_MASK, /* LVT0-1 */
137 LVT_MASK /* LVTERR */
138};
139
140static int find_highest_vector(void *bitmap)
141{
142 u32 *word = bitmap;
143 int word_offset = MAX_APIC_VECTOR >> 5;
144
145 while ((word_offset != 0) && (word[(--word_offset) << 2] == 0))
146 continue;
147
148 if (likely(!word_offset && !word[0]))
149 return -1;
150 else
151 return fls(word[word_offset << 2]) - 1 + (word_offset << 5);
152}
153
154static inline int apic_test_and_set_irr(int vec, struct kvm_lapic *apic)
155{
156 return apic_test_and_set_vector(vec, apic->regs + APIC_IRR);
157}
158
159static inline void apic_clear_irr(int vec, struct kvm_lapic *apic)
160{
161 apic_clear_vector(vec, apic->regs + APIC_IRR);
162}
163
164static inline int apic_find_highest_irr(struct kvm_lapic *apic)
165{
166 int result;
167
168 result = find_highest_vector(apic->regs + APIC_IRR);
169 ASSERT(result == -1 || result >= 16);
170
171 return result;
172}
173
6e5d865c
YS
174int kvm_lapic_find_highest_irr(struct kvm_vcpu *vcpu)
175{
ad312c7c 176 struct kvm_lapic *apic = vcpu->arch.apic;
6e5d865c
YS
177 int highest_irr;
178
179 if (!apic)
180 return 0;
181 highest_irr = apic_find_highest_irr(apic);
182
183 return highest_irr;
184}
185EXPORT_SYMBOL_GPL(kvm_lapic_find_highest_irr);
186
8be5453f 187int kvm_apic_set_irq(struct kvm_vcpu *vcpu, u8 vec, u8 trig)
97222cc8 188{
ad312c7c 189 struct kvm_lapic *apic = vcpu->arch.apic;
8be5453f 190
97222cc8
ED
191 if (!apic_test_and_set_irr(vec, apic)) {
192 /* a new pending irq is set in IRR */
193 if (trig)
194 apic_set_vector(vec, apic->regs + APIC_TMR);
195 else
196 apic_clear_vector(vec, apic->regs + APIC_TMR);
197 kvm_vcpu_kick(apic->vcpu);
198 return 1;
199 }
200 return 0;
201}
202
203static inline int apic_find_highest_isr(struct kvm_lapic *apic)
204{
205 int result;
206
207 result = find_highest_vector(apic->regs + APIC_ISR);
208 ASSERT(result == -1 || result >= 16);
209
210 return result;
211}
212
213static void apic_update_ppr(struct kvm_lapic *apic)
214{
215 u32 tpr, isrv, ppr;
216 int isr;
217
218 tpr = apic_get_reg(apic, APIC_TASKPRI);
219 isr = apic_find_highest_isr(apic);
220 isrv = (isr != -1) ? isr : 0;
221
222 if ((tpr & 0xf0) >= (isrv & 0xf0))
223 ppr = tpr & 0xff;
224 else
225 ppr = isrv & 0xf0;
226
227 apic_debug("vlapic %p, ppr 0x%x, isr 0x%x, isrv 0x%x",
228 apic, ppr, isr, isrv);
229
230 apic_set_reg(apic, APIC_PROCPRI, ppr);
231}
232
233static void apic_set_tpr(struct kvm_lapic *apic, u32 tpr)
234{
235 apic_set_reg(apic, APIC_TASKPRI, tpr);
236 apic_update_ppr(apic);
237}
238
239int kvm_apic_match_physical_addr(struct kvm_lapic *apic, u16 dest)
240{
241 return kvm_apic_id(apic) == dest;
242}
243
244int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
245{
246 int result = 0;
247 u8 logical_id;
248
249 logical_id = GET_APIC_LOGICAL_ID(apic_get_reg(apic, APIC_LDR));
250
251 switch (apic_get_reg(apic, APIC_DFR)) {
252 case APIC_DFR_FLAT:
253 if (logical_id & mda)
254 result = 1;
255 break;
256 case APIC_DFR_CLUSTER:
257 if (((logical_id >> 4) == (mda >> 0x4))
258 && (logical_id & mda & 0xf))
259 result = 1;
260 break;
261 default:
262 printk(KERN_WARNING "Bad DFR vcpu %d: %08x\n",
263 apic->vcpu->vcpu_id, apic_get_reg(apic, APIC_DFR));
264 break;
265 }
266
267 return result;
268}
269
270static int apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
271 int short_hand, int dest, int dest_mode)
272{
273 int result = 0;
ad312c7c 274 struct kvm_lapic *target = vcpu->arch.apic;
97222cc8
ED
275
276 apic_debug("target %p, source %p, dest 0x%x, "
277 "dest_mode 0x%x, short_hand 0x%x",
278 target, source, dest, dest_mode, short_hand);
279
280 ASSERT(!target);
281 switch (short_hand) {
282 case APIC_DEST_NOSHORT:
283 if (dest_mode == 0) {
284 /* Physical mode. */
285 if ((dest == 0xFF) || (dest == kvm_apic_id(target)))
286 result = 1;
287 } else
288 /* Logical mode. */
289 result = kvm_apic_match_logical_addr(target, dest);
290 break;
291 case APIC_DEST_SELF:
292 if (target == source)
293 result = 1;
294 break;
295 case APIC_DEST_ALLINC:
296 result = 1;
297 break;
298 case APIC_DEST_ALLBUT:
299 if (target != source)
300 result = 1;
301 break;
302 default:
303 printk(KERN_WARNING "Bad dest shorthand value %x\n",
304 short_hand);
305 break;
306 }
307
308 return result;
309}
310
311/*
312 * Add a pending IRQ into lapic.
313 * Return 1 if successfully added and 0 if discarded.
314 */
315static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
316 int vector, int level, int trig_mode)
317{
c5ec1534
HQ
318 int orig_irr, result = 0;
319 struct kvm_vcpu *vcpu = apic->vcpu;
97222cc8
ED
320
321 switch (delivery_mode) {
322 case APIC_DM_FIXED:
323 case APIC_DM_LOWEST:
324 /* FIXME add logic for vcpu on reset */
325 if (unlikely(!apic_enabled(apic)))
326 break;
327
1b9778da
ED
328 orig_irr = apic_test_and_set_irr(vector, apic);
329 if (orig_irr && trig_mode) {
97222cc8
ED
330 apic_debug("level trig mode repeatedly for vector %d",
331 vector);
332 break;
333 }
334
335 if (trig_mode) {
336 apic_debug("level trig mode for vector %d", vector);
337 apic_set_vector(vector, apic->regs + APIC_TMR);
338 } else
339 apic_clear_vector(vector, apic->regs + APIC_TMR);
340
a4535290 341 if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE)
c5ec1534 342 kvm_vcpu_kick(vcpu);
a4535290
AK
343 else if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED) {
344 vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
c5ec1534
HQ
345 if (waitqueue_active(&vcpu->wq))
346 wake_up_interruptible(&vcpu->wq);
347 }
97222cc8 348
1b9778da 349 result = (orig_irr == 0);
97222cc8
ED
350 break;
351
352 case APIC_DM_REMRD:
353 printk(KERN_DEBUG "Ignoring delivery mode 3\n");
354 break;
355
356 case APIC_DM_SMI:
357 printk(KERN_DEBUG "Ignoring guest SMI\n");
358 break;
359 case APIC_DM_NMI:
360 printk(KERN_DEBUG "Ignoring guest NMI\n");
361 break;
362
363 case APIC_DM_INIT:
c5ec1534 364 if (level) {
a4535290 365 if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE)
c5ec1534
HQ
366 printk(KERN_DEBUG
367 "INIT on a runnable vcpu %d\n",
368 vcpu->vcpu_id);
a4535290 369 vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
c5ec1534
HQ
370 kvm_vcpu_kick(vcpu);
371 } else {
372 printk(KERN_DEBUG
373 "Ignoring de-assert INIT to vcpu %d\n",
374 vcpu->vcpu_id);
375 }
376
97222cc8
ED
377 break;
378
379 case APIC_DM_STARTUP:
c5ec1534
HQ
380 printk(KERN_DEBUG "SIPI to vcpu %d vector 0x%02x\n",
381 vcpu->vcpu_id, vector);
a4535290 382 if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) {
ad312c7c 383 vcpu->arch.sipi_vector = vector;
a4535290 384 vcpu->arch.mp_state = KVM_MP_STATE_SIPI_RECEIVED;
c5ec1534
HQ
385 if (waitqueue_active(&vcpu->wq))
386 wake_up_interruptible(&vcpu->wq);
387 }
97222cc8
ED
388 break;
389
390 default:
391 printk(KERN_ERR "TODO: unsupported delivery mode %x\n",
392 delivery_mode);
393 break;
394 }
395 return result;
396}
397
8be5453f 398static struct kvm_lapic *kvm_apic_round_robin(struct kvm *kvm, u8 vector,
97222cc8
ED
399 unsigned long bitmap)
400{
932f72ad
HQ
401 int last;
402 int next;
e4d47f40 403 struct kvm_lapic *apic = NULL;
932f72ad 404
bfc6d222 405 last = kvm->arch.round_robin_prev_vcpu;
932f72ad
HQ
406 next = last;
407
408 do {
409 if (++next == KVM_MAX_VCPUS)
410 next = 0;
411 if (kvm->vcpus[next] == NULL || !test_bit(next, &bitmap))
412 continue;
ad312c7c 413 apic = kvm->vcpus[next]->arch.apic;
932f72ad
HQ
414 if (apic && apic_enabled(apic))
415 break;
416 apic = NULL;
417 } while (next != last);
bfc6d222 418 kvm->arch.round_robin_prev_vcpu = next;
932f72ad 419
e4d47f40
QH
420 if (!apic)
421 printk(KERN_DEBUG "vcpu not ready for apic_round_robin\n");
97222cc8 422
932f72ad 423 return apic;
97222cc8
ED
424}
425
8be5453f
ZX
426struct kvm_vcpu *kvm_get_lowest_prio_vcpu(struct kvm *kvm, u8 vector,
427 unsigned long bitmap)
428{
429 struct kvm_lapic *apic;
430
431 apic = kvm_apic_round_robin(kvm, vector, bitmap);
432 if (apic)
433 return apic->vcpu;
434 return NULL;
435}
436
97222cc8
ED
437static void apic_set_eoi(struct kvm_lapic *apic)
438{
439 int vector = apic_find_highest_isr(apic);
440
441 /*
442 * Not every write EOI will has corresponding ISR,
443 * one example is when Kernel check timer on setup_IO_APIC
444 */
445 if (vector == -1)
446 return;
447
448 apic_clear_vector(vector, apic->regs + APIC_ISR);
449 apic_update_ppr(apic);
450
451 if (apic_test_and_clear_vector(vector, apic->regs + APIC_TMR))
452 kvm_ioapic_update_eoi(apic->vcpu->kvm, vector);
453}
454
455static void apic_send_ipi(struct kvm_lapic *apic)
456{
457 u32 icr_low = apic_get_reg(apic, APIC_ICR);
458 u32 icr_high = apic_get_reg(apic, APIC_ICR2);
459
460 unsigned int dest = GET_APIC_DEST_FIELD(icr_high);
461 unsigned int short_hand = icr_low & APIC_SHORT_MASK;
462 unsigned int trig_mode = icr_low & APIC_INT_LEVELTRIG;
463 unsigned int level = icr_low & APIC_INT_ASSERT;
464 unsigned int dest_mode = icr_low & APIC_DEST_MASK;
465 unsigned int delivery_mode = icr_low & APIC_MODE_MASK;
466 unsigned int vector = icr_low & APIC_VECTOR_MASK;
467
8be5453f 468 struct kvm_vcpu *target;
97222cc8
ED
469 struct kvm_vcpu *vcpu;
470 unsigned long lpr_map = 0;
471 int i;
472
473 apic_debug("icr_high 0x%x, icr_low 0x%x, "
474 "short_hand 0x%x, dest 0x%x, trig_mode 0x%x, level 0x%x, "
475 "dest_mode 0x%x, delivery_mode 0x%x, vector 0x%x\n",
476 icr_high, icr_low, short_hand, dest,
477 trig_mode, level, dest_mode, delivery_mode, vector);
478
479 for (i = 0; i < KVM_MAX_VCPUS; i++) {
480 vcpu = apic->vcpu->kvm->vcpus[i];
481 if (!vcpu)
482 continue;
483
ad312c7c 484 if (vcpu->arch.apic &&
97222cc8
ED
485 apic_match_dest(vcpu, apic, short_hand, dest, dest_mode)) {
486 if (delivery_mode == APIC_DM_LOWEST)
487 set_bit(vcpu->vcpu_id, &lpr_map);
488 else
ad312c7c 489 __apic_accept_irq(vcpu->arch.apic, delivery_mode,
97222cc8
ED
490 vector, level, trig_mode);
491 }
492 }
493
494 if (delivery_mode == APIC_DM_LOWEST) {
8be5453f 495 target = kvm_get_lowest_prio_vcpu(vcpu->kvm, vector, lpr_map);
97222cc8 496 if (target != NULL)
ad312c7c 497 __apic_accept_irq(target->arch.apic, delivery_mode,
97222cc8
ED
498 vector, level, trig_mode);
499 }
500}
501
502static u32 apic_get_tmcct(struct kvm_lapic *apic)
503{
9da8f4e8
KP
504 u64 counter_passed;
505 ktime_t passed, now;
506 u32 tmcct;
97222cc8
ED
507
508 ASSERT(apic != NULL);
509
9da8f4e8
KP
510 now = apic->timer.dev.base->get_time();
511 tmcct = apic_get_reg(apic, APIC_TMICT);
512
513 /* if initial count is 0, current count should also be 0 */
514 if (tmcct == 0)
515 return 0;
516
97222cc8
ED
517 if (unlikely(ktime_to_ns(now) <=
518 ktime_to_ns(apic->timer.last_update))) {
519 /* Wrap around */
520 passed = ktime_add(( {
521 (ktime_t) {
522 .tv64 = KTIME_MAX -
523 (apic->timer.last_update).tv64}; }
524 ), now);
525 apic_debug("time elapsed\n");
526 } else
527 passed = ktime_sub(now, apic->timer.last_update);
528
6f6d6a1a
RZ
529 counter_passed = div64_u64(ktime_to_ns(passed),
530 (APIC_BUS_CYCLE_NS * apic->timer.divide_count));
97222cc8 531
9da8f4e8
KP
532 if (counter_passed > tmcct) {
533 if (unlikely(!apic_lvtt_period(apic))) {
534 /* one-shot timers stick at 0 until reset */
97222cc8 535 tmcct = 0;
9da8f4e8
KP
536 } else {
537 /*
538 * periodic timers reset to APIC_TMICT when they
539 * hit 0. The while loop simulates this happening N
540 * times. (counter_passed %= tmcct) would also work,
541 * but might be slower or not work on 32-bit??
542 */
543 while (counter_passed > tmcct)
544 counter_passed -= tmcct;
545 tmcct -= counter_passed;
546 }
547 } else {
548 tmcct -= counter_passed;
97222cc8
ED
549 }
550
551 return tmcct;
552}
553
b209749f
AK
554static 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 set_bit(KVM_REQ_REPORT_TPR_ACCESS, &vcpu->requests);
560 kvm_x86_ops->cache_regs(vcpu);
561 run->tpr_access.rip = vcpu->arch.rip;
562 run->tpr_access.is_write = write;
563}
564
565static inline void report_tpr_access(struct kvm_lapic *apic, bool write)
566{
567 if (apic->vcpu->arch.tpr_access_reporting)
568 __report_tpr_access(apic, write);
569}
570
97222cc8
ED
571static u32 __apic_read(struct kvm_lapic *apic, unsigned int offset)
572{
573 u32 val = 0;
574
575 if (offset >= LAPIC_MMIO_LENGTH)
576 return 0;
577
578 switch (offset) {
579 case APIC_ARBPRI:
580 printk(KERN_WARNING "Access APIC ARBPRI register "
581 "which is for P6\n");
582 break;
583
584 case APIC_TMCCT: /* Timer CCR */
585 val = apic_get_tmcct(apic);
586 break;
587
b209749f
AK
588 case APIC_TASKPRI:
589 report_tpr_access(apic, false);
590 /* fall thru */
97222cc8 591 default:
6e5d865c 592 apic_update_ppr(apic);
97222cc8
ED
593 val = apic_get_reg(apic, offset);
594 break;
595 }
596
597 return val;
598}
599
600static void apic_mmio_read(struct kvm_io_device *this,
601 gpa_t address, int len, void *data)
602{
603 struct kvm_lapic *apic = (struct kvm_lapic *)this->private;
604 unsigned int offset = address - apic->base_address;
605 unsigned char alignment = offset & 0xf;
606 u32 result;
607
608 if ((alignment + len) > 4) {
609 printk(KERN_ERR "KVM_APIC_READ: alignment error %lx %d",
610 (unsigned long)address, len);
611 return;
612 }
613 result = __apic_read(apic, offset & ~0xf);
614
615 switch (len) {
616 case 1:
617 case 2:
618 case 4:
619 memcpy(data, (char *)&result + alignment, len);
620 break;
621 default:
622 printk(KERN_ERR "Local APIC read with len = %x, "
623 "should be 1,2, or 4 instead\n", len);
624 break;
625 }
626}
627
628static void update_divide_count(struct kvm_lapic *apic)
629{
630 u32 tmp1, tmp2, tdcr;
631
632 tdcr = apic_get_reg(apic, APIC_TDCR);
633 tmp1 = tdcr & 0xf;
634 tmp2 = ((tmp1 & 0x3) | ((tmp1 & 0x8) >> 1)) + 1;
635 apic->timer.divide_count = 0x1 << (tmp2 & 0x7);
636
637 apic_debug("timer divide count is 0x%x\n",
638 apic->timer.divide_count);
639}
640
641static void start_apic_timer(struct kvm_lapic *apic)
642{
643 ktime_t now = apic->timer.dev.base->get_time();
644
645 apic->timer.last_update = now;
646
647 apic->timer.period = apic_get_reg(apic, APIC_TMICT) *
648 APIC_BUS_CYCLE_NS * apic->timer.divide_count;
649 atomic_set(&apic->timer.pending, 0);
0b975a3c
AK
650
651 if (!apic->timer.period)
652 return;
653
97222cc8
ED
654 hrtimer_start(&apic->timer.dev,
655 ktime_add_ns(now, apic->timer.period),
656 HRTIMER_MODE_ABS);
657
658 apic_debug("%s: bus cycle is %" PRId64 "ns, now 0x%016"
659 PRIx64 ", "
660 "timer initial count 0x%x, period %lldns, "
b8688d51 661 "expire @ 0x%016" PRIx64 ".\n", __func__,
97222cc8
ED
662 APIC_BUS_CYCLE_NS, ktime_to_ns(now),
663 apic_get_reg(apic, APIC_TMICT),
664 apic->timer.period,
665 ktime_to_ns(ktime_add_ns(now,
666 apic->timer.period)));
667}
668
669static void apic_mmio_write(struct kvm_io_device *this,
670 gpa_t address, int len, const void *data)
671{
672 struct kvm_lapic *apic = (struct kvm_lapic *)this->private;
673 unsigned int offset = address - apic->base_address;
674 unsigned char alignment = offset & 0xf;
675 u32 val;
676
677 /*
678 * APIC register must be aligned on 128-bits boundary.
679 * 32/64/128 bits registers must be accessed thru 32 bits.
680 * Refer SDM 8.4.1
681 */
682 if (len != 4 || alignment) {
683 if (printk_ratelimit())
684 printk(KERN_ERR "apic write: bad size=%d %lx\n",
685 len, (long)address);
686 return;
687 }
688
689 val = *(u32 *) data;
690
691 /* too common printing */
692 if (offset != APIC_EOI)
693 apic_debug("%s: offset 0x%x with length 0x%x, and value is "
b8688d51 694 "0x%x\n", __func__, offset, len, val);
97222cc8
ED
695
696 offset &= 0xff0;
697
698 switch (offset) {
699 case APIC_ID: /* Local APIC ID */
700 apic_set_reg(apic, APIC_ID, val);
701 break;
702
703 case APIC_TASKPRI:
b209749f 704 report_tpr_access(apic, true);
97222cc8
ED
705 apic_set_tpr(apic, val & 0xff);
706 break;
707
708 case APIC_EOI:
709 apic_set_eoi(apic);
710 break;
711
712 case APIC_LDR:
713 apic_set_reg(apic, APIC_LDR, val & APIC_LDR_MASK);
714 break;
715
716 case APIC_DFR:
717 apic_set_reg(apic, APIC_DFR, val | 0x0FFFFFFF);
718 break;
719
720 case APIC_SPIV:
721 apic_set_reg(apic, APIC_SPIV, val & 0x3ff);
722 if (!(val & APIC_SPIV_APIC_ENABLED)) {
723 int i;
724 u32 lvt_val;
725
726 for (i = 0; i < APIC_LVT_NUM; i++) {
727 lvt_val = apic_get_reg(apic,
728 APIC_LVTT + 0x10 * i);
729 apic_set_reg(apic, APIC_LVTT + 0x10 * i,
730 lvt_val | APIC_LVT_MASKED);
731 }
732 atomic_set(&apic->timer.pending, 0);
733
734 }
735 break;
736
737 case APIC_ICR:
738 /* No delay here, so we always clear the pending bit */
739 apic_set_reg(apic, APIC_ICR, val & ~(1 << 12));
740 apic_send_ipi(apic);
741 break;
742
743 case APIC_ICR2:
744 apic_set_reg(apic, APIC_ICR2, val & 0xff000000);
745 break;
746
747 case APIC_LVTT:
748 case APIC_LVTTHMR:
749 case APIC_LVTPC:
750 case APIC_LVT0:
751 case APIC_LVT1:
752 case APIC_LVTERR:
753 /* TODO: Check vector */
754 if (!apic_sw_enabled(apic))
755 val |= APIC_LVT_MASKED;
756
757 val &= apic_lvt_mask[(offset - APIC_LVTT) >> 4];
758 apic_set_reg(apic, offset, val);
759
760 break;
761
762 case APIC_TMICT:
763 hrtimer_cancel(&apic->timer.dev);
764 apic_set_reg(apic, APIC_TMICT, val);
765 start_apic_timer(apic);
766 return;
767
768 case APIC_TDCR:
769 if (val & 4)
770 printk(KERN_ERR "KVM_WRITE:TDCR %x\n", val);
771 apic_set_reg(apic, APIC_TDCR, val);
772 update_divide_count(apic);
773 break;
774
775 default:
776 apic_debug("Local APIC Write to read-only register %x\n",
777 offset);
778 break;
779 }
780
781}
782
783static int apic_mmio_range(struct kvm_io_device *this, gpa_t addr)
784{
785 struct kvm_lapic *apic = (struct kvm_lapic *)this->private;
786 int ret = 0;
787
788
789 if (apic_hw_enabled(apic) &&
790 (addr >= apic->base_address) &&
791 (addr < (apic->base_address + LAPIC_MMIO_LENGTH)))
792 ret = 1;
793
794 return ret;
795}
796
d589444e 797void kvm_free_lapic(struct kvm_vcpu *vcpu)
97222cc8 798{
ad312c7c 799 if (!vcpu->arch.apic)
97222cc8
ED
800 return;
801
ad312c7c 802 hrtimer_cancel(&vcpu->arch.apic->timer.dev);
97222cc8 803
ad312c7c
ZX
804 if (vcpu->arch.apic->regs_page)
805 __free_page(vcpu->arch.apic->regs_page);
97222cc8 806
ad312c7c 807 kfree(vcpu->arch.apic);
97222cc8
ED
808}
809
810/*
811 *----------------------------------------------------------------------
812 * LAPIC interface
813 *----------------------------------------------------------------------
814 */
815
816void kvm_lapic_set_tpr(struct kvm_vcpu *vcpu, unsigned long cr8)
817{
ad312c7c 818 struct kvm_lapic *apic = vcpu->arch.apic;
97222cc8
ED
819
820 if (!apic)
821 return;
b93463aa
AK
822 apic_set_tpr(apic, ((cr8 & 0x0f) << 4)
823 | (apic_get_reg(apic, APIC_TASKPRI) & 4));
97222cc8 824}
ec7cf690 825EXPORT_SYMBOL_GPL(kvm_lapic_set_tpr);
97222cc8
ED
826
827u64 kvm_lapic_get_cr8(struct kvm_vcpu *vcpu)
828{
ad312c7c 829 struct kvm_lapic *apic = vcpu->arch.apic;
97222cc8
ED
830 u64 tpr;
831
832 if (!apic)
833 return 0;
834 tpr = (u64) apic_get_reg(apic, APIC_TASKPRI);
835
836 return (tpr & 0xf0) >> 4;
837}
6e5d865c 838EXPORT_SYMBOL_GPL(kvm_lapic_get_cr8);
97222cc8
ED
839
840void kvm_lapic_set_base(struct kvm_vcpu *vcpu, u64 value)
841{
ad312c7c 842 struct kvm_lapic *apic = vcpu->arch.apic;
97222cc8
ED
843
844 if (!apic) {
845 value |= MSR_IA32_APICBASE_BSP;
ad312c7c 846 vcpu->arch.apic_base = value;
97222cc8
ED
847 return;
848 }
849 if (apic->vcpu->vcpu_id)
850 value &= ~MSR_IA32_APICBASE_BSP;
851
ad312c7c
ZX
852 vcpu->arch.apic_base = value;
853 apic->base_address = apic->vcpu->arch.apic_base &
97222cc8
ED
854 MSR_IA32_APICBASE_BASE;
855
856 /* with FSB delivery interrupt, we can restart APIC functionality */
857 apic_debug("apic base msr is 0x%016" PRIx64 ", and base address is "
ad312c7c 858 "0x%lx.\n", apic->vcpu->arch.apic_base, apic->base_address);
97222cc8
ED
859
860}
861
862u64 kvm_lapic_get_base(struct kvm_vcpu *vcpu)
863{
ad312c7c 864 return vcpu->arch.apic_base;
97222cc8
ED
865}
866EXPORT_SYMBOL_GPL(kvm_lapic_get_base);
867
c5ec1534 868void kvm_lapic_reset(struct kvm_vcpu *vcpu)
97222cc8
ED
869{
870 struct kvm_lapic *apic;
871 int i;
872
b8688d51 873 apic_debug("%s\n", __func__);
97222cc8
ED
874
875 ASSERT(vcpu);
ad312c7c 876 apic = vcpu->arch.apic;
97222cc8
ED
877 ASSERT(apic != NULL);
878
879 /* Stop the timer in case it's a reset to an active apic */
880 hrtimer_cancel(&apic->timer.dev);
881
882 apic_set_reg(apic, APIC_ID, vcpu->vcpu_id << 24);
883 apic_set_reg(apic, APIC_LVR, APIC_VERSION);
884
885 for (i = 0; i < APIC_LVT_NUM; i++)
886 apic_set_reg(apic, APIC_LVTT + 0x10 * i, APIC_LVT_MASKED);
40487c68
QH
887 apic_set_reg(apic, APIC_LVT0,
888 SET_APIC_DELIVERY_MODE(0, APIC_MODE_EXTINT));
97222cc8
ED
889
890 apic_set_reg(apic, APIC_DFR, 0xffffffffU);
891 apic_set_reg(apic, APIC_SPIV, 0xff);
892 apic_set_reg(apic, APIC_TASKPRI, 0);
893 apic_set_reg(apic, APIC_LDR, 0);
894 apic_set_reg(apic, APIC_ESR, 0);
895 apic_set_reg(apic, APIC_ICR, 0);
896 apic_set_reg(apic, APIC_ICR2, 0);
897 apic_set_reg(apic, APIC_TDCR, 0);
898 apic_set_reg(apic, APIC_TMICT, 0);
899 for (i = 0; i < 8; i++) {
900 apic_set_reg(apic, APIC_IRR + 0x10 * i, 0);
901 apic_set_reg(apic, APIC_ISR + 0x10 * i, 0);
902 apic_set_reg(apic, APIC_TMR + 0x10 * i, 0);
903 }
b33ac88b 904 update_divide_count(apic);
97222cc8
ED
905 atomic_set(&apic->timer.pending, 0);
906 if (vcpu->vcpu_id == 0)
ad312c7c 907 vcpu->arch.apic_base |= MSR_IA32_APICBASE_BSP;
97222cc8
ED
908 apic_update_ppr(apic);
909
910 apic_debug(KERN_INFO "%s: vcpu=%p, id=%d, base_msr="
b8688d51 911 "0x%016" PRIx64 ", base_address=0x%0lx.\n", __func__,
97222cc8 912 vcpu, kvm_apic_id(apic),
ad312c7c 913 vcpu->arch.apic_base, apic->base_address);
97222cc8 914}
c5ec1534 915EXPORT_SYMBOL_GPL(kvm_lapic_reset);
97222cc8
ED
916
917int kvm_lapic_enabled(struct kvm_vcpu *vcpu)
918{
ad312c7c 919 struct kvm_lapic *apic = vcpu->arch.apic;
97222cc8
ED
920 int ret = 0;
921
922 if (!apic)
923 return 0;
924 ret = apic_enabled(apic);
925
926 return ret;
927}
6e5d865c 928EXPORT_SYMBOL_GPL(kvm_lapic_enabled);
97222cc8
ED
929
930/*
931 *----------------------------------------------------------------------
932 * timer interface
933 *----------------------------------------------------------------------
934 */
1b9778da
ED
935
936/* TODO: make sure __apic_timer_fn runs in current pCPU */
97222cc8
ED
937static int __apic_timer_fn(struct kvm_lapic *apic)
938{
97222cc8 939 int result = 0;
1b9778da 940 wait_queue_head_t *q = &apic->vcpu->wq;
97222cc8 941
97222cc8 942 atomic_inc(&apic->timer.pending);
d77c26fc 943 if (waitqueue_active(q)) {
a4535290 944 apic->vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
1b9778da 945 wake_up_interruptible(q);
c5ec1534 946 }
97222cc8 947 if (apic_lvtt_period(apic)) {
97222cc8
ED
948 result = 1;
949 apic->timer.dev.expires = ktime_add_ns(
950 apic->timer.dev.expires,
951 apic->timer.period);
952 }
97222cc8
ED
953 return result;
954}
955
3d80840d
MT
956int apic_has_pending_timer(struct kvm_vcpu *vcpu)
957{
958 struct kvm_lapic *lapic = vcpu->arch.apic;
959
960 if (lapic)
961 return atomic_read(&lapic->timer.pending);
962
963 return 0;
964}
965
1b9778da
ED
966static int __inject_apic_timer_irq(struct kvm_lapic *apic)
967{
968 int vector;
969
970 vector = apic_lvt_vector(apic, APIC_LVTT);
971 return __apic_accept_irq(apic, APIC_DM_FIXED, vector, 1, 0);
972}
973
97222cc8
ED
974static enum hrtimer_restart apic_timer_fn(struct hrtimer *data)
975{
976 struct kvm_lapic *apic;
977 int restart_timer = 0;
978
979 apic = container_of(data, struct kvm_lapic, timer.dev);
980
981 restart_timer = __apic_timer_fn(apic);
982
983 if (restart_timer)
984 return HRTIMER_RESTART;
985 else
986 return HRTIMER_NORESTART;
987}
988
989int kvm_create_lapic(struct kvm_vcpu *vcpu)
990{
991 struct kvm_lapic *apic;
992
993 ASSERT(vcpu != NULL);
994 apic_debug("apic_init %d\n", vcpu->vcpu_id);
995
996 apic = kzalloc(sizeof(*apic), GFP_KERNEL);
997 if (!apic)
998 goto nomem;
999
ad312c7c 1000 vcpu->arch.apic = apic;
97222cc8
ED
1001
1002 apic->regs_page = alloc_page(GFP_KERNEL);
1003 if (apic->regs_page == NULL) {
1004 printk(KERN_ERR "malloc apic regs error for vcpu %x\n",
1005 vcpu->vcpu_id);
d589444e 1006 goto nomem_free_apic;
97222cc8
ED
1007 }
1008 apic->regs = page_address(apic->regs_page);
1009 memset(apic->regs, 0, PAGE_SIZE);
1010 apic->vcpu = vcpu;
1011
1012 hrtimer_init(&apic->timer.dev, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
1013 apic->timer.dev.function = apic_timer_fn;
1014 apic->base_address = APIC_DEFAULT_PHYS_BASE;
ad312c7c 1015 vcpu->arch.apic_base = APIC_DEFAULT_PHYS_BASE;
97222cc8 1016
c5ec1534 1017 kvm_lapic_reset(vcpu);
97222cc8
ED
1018 apic->dev.read = apic_mmio_read;
1019 apic->dev.write = apic_mmio_write;
1020 apic->dev.in_range = apic_mmio_range;
1021 apic->dev.private = apic;
1022
1023 return 0;
d589444e
RR
1024nomem_free_apic:
1025 kfree(apic);
97222cc8 1026nomem:
97222cc8
ED
1027 return -ENOMEM;
1028}
1029EXPORT_SYMBOL_GPL(kvm_create_lapic);
1030
1031int kvm_apic_has_interrupt(struct kvm_vcpu *vcpu)
1032{
ad312c7c 1033 struct kvm_lapic *apic = vcpu->arch.apic;
97222cc8
ED
1034 int highest_irr;
1035
1036 if (!apic || !apic_enabled(apic))
1037 return -1;
1038
6e5d865c 1039 apic_update_ppr(apic);
97222cc8
ED
1040 highest_irr = apic_find_highest_irr(apic);
1041 if ((highest_irr == -1) ||
1042 ((highest_irr & 0xF0) <= apic_get_reg(apic, APIC_PROCPRI)))
1043 return -1;
1044 return highest_irr;
1045}
1046
40487c68
QH
1047int kvm_apic_accept_pic_intr(struct kvm_vcpu *vcpu)
1048{
ad312c7c 1049 u32 lvt0 = apic_get_reg(vcpu->arch.apic, APIC_LVT0);
40487c68
QH
1050 int r = 0;
1051
1052 if (vcpu->vcpu_id == 0) {
ad312c7c 1053 if (!apic_hw_enabled(vcpu->arch.apic))
40487c68
QH
1054 r = 1;
1055 if ((lvt0 & APIC_LVT_MASKED) == 0 &&
1056 GET_APIC_DELIVERY_MODE(lvt0) == APIC_MODE_EXTINT)
1057 r = 1;
1058 }
1059 return r;
1060}
1061
1b9778da
ED
1062void kvm_inject_apic_timer_irqs(struct kvm_vcpu *vcpu)
1063{
ad312c7c 1064 struct kvm_lapic *apic = vcpu->arch.apic;
1b9778da
ED
1065
1066 if (apic && apic_lvt_enabled(apic, APIC_LVTT) &&
1067 atomic_read(&apic->timer.pending) > 0) {
1068 if (__inject_apic_timer_irq(apic))
1069 atomic_dec(&apic->timer.pending);
1070 }
1071}
1072
1073void kvm_apic_timer_intr_post(struct kvm_vcpu *vcpu, int vec)
1074{
ad312c7c 1075 struct kvm_lapic *apic = vcpu->arch.apic;
1b9778da
ED
1076
1077 if (apic && apic_lvt_vector(apic, APIC_LVTT) == vec)
1078 apic->timer.last_update = ktime_add_ns(
1079 apic->timer.last_update,
1080 apic->timer.period);
1081}
1082
97222cc8
ED
1083int kvm_get_apic_interrupt(struct kvm_vcpu *vcpu)
1084{
1085 int vector = kvm_apic_has_interrupt(vcpu);
ad312c7c 1086 struct kvm_lapic *apic = vcpu->arch.apic;
97222cc8
ED
1087
1088 if (vector == -1)
1089 return -1;
1090
1091 apic_set_vector(vector, apic->regs + APIC_ISR);
1092 apic_update_ppr(apic);
1093 apic_clear_irr(vector, apic);
1094 return vector;
1095}
96ad2cc6
ED
1096
1097void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu)
1098{
ad312c7c 1099 struct kvm_lapic *apic = vcpu->arch.apic;
96ad2cc6 1100
ad312c7c 1101 apic->base_address = vcpu->arch.apic_base &
96ad2cc6
ED
1102 MSR_IA32_APICBASE_BASE;
1103 apic_set_reg(apic, APIC_LVR, APIC_VERSION);
1104 apic_update_ppr(apic);
1105 hrtimer_cancel(&apic->timer.dev);
1106 update_divide_count(apic);
1107 start_apic_timer(apic);
1108}
a3d7f85f 1109
2f52d58c 1110void __kvm_migrate_apic_timer(struct kvm_vcpu *vcpu)
a3d7f85f 1111{
ad312c7c 1112 struct kvm_lapic *apic = vcpu->arch.apic;
a3d7f85f
ED
1113 struct hrtimer *timer;
1114
1115 if (!apic)
1116 return;
1117
1118 timer = &apic->timer.dev;
1119 if (hrtimer_cancel(timer))
1120 hrtimer_start(timer, timer->expires, HRTIMER_MODE_ABS);
1121}
b93463aa
AK
1122
1123void kvm_lapic_sync_from_vapic(struct kvm_vcpu *vcpu)
1124{
1125 u32 data;
1126 void *vapic;
1127
1128 if (!irqchip_in_kernel(vcpu->kvm) || !vcpu->arch.apic->vapic_addr)
1129 return;
1130
1131 vapic = kmap_atomic(vcpu->arch.apic->vapic_page, KM_USER0);
1132 data = *(u32 *)(vapic + offset_in_page(vcpu->arch.apic->vapic_addr));
1133 kunmap_atomic(vapic, KM_USER0);
1134
1135 apic_set_tpr(vcpu->arch.apic, data & 0xff);
1136}
1137
1138void kvm_lapic_sync_to_vapic(struct kvm_vcpu *vcpu)
1139{
1140 u32 data, tpr;
1141 int max_irr, max_isr;
1142 struct kvm_lapic *apic;
1143 void *vapic;
1144
1145 if (!irqchip_in_kernel(vcpu->kvm) || !vcpu->arch.apic->vapic_addr)
1146 return;
1147
1148 apic = vcpu->arch.apic;
1149 tpr = apic_get_reg(apic, APIC_TASKPRI) & 0xff;
1150 max_irr = apic_find_highest_irr(apic);
1151 if (max_irr < 0)
1152 max_irr = 0;
1153 max_isr = apic_find_highest_isr(apic);
1154 if (max_isr < 0)
1155 max_isr = 0;
1156 data = (tpr & 0xff) | ((max_isr & 0xf0) << 8) | (max_irr << 24);
1157
1158 vapic = kmap_atomic(vcpu->arch.apic->vapic_page, KM_USER0);
1159 *(u32 *)(vapic + offset_in_page(vcpu->arch.apic->vapic_addr)) = data;
1160 kunmap_atomic(vapic, KM_USER0);
1161}
1162
1163void kvm_lapic_set_vapic_addr(struct kvm_vcpu *vcpu, gpa_t vapic_addr)
1164{
1165 if (!irqchip_in_kernel(vcpu->kvm))
1166 return;
1167
1168 vcpu->arch.apic->vapic_addr = vapic_addr;
1169}