2 * Copyright (C) 2001 MandrakeSoft S.A.
3 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
8 * http://www.linux-mandrake.com/
9 * http://www.mandrakesoft.com/
11 * This library is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2 of the License, or (at your option) any later version.
16 * This library is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with this library; if not, write to the Free Software
23 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
25 * Yunhong Jiang <yunhong.jiang@intel.com>
26 * Yaozu (Eddie) Dong <eddie.dong@intel.com>
27 * Based on Xen 3.1 code.
30 #include <linux/kvm_host.h>
31 #include <linux/kvm.h>
33 #include <linux/highmem.h>
34 #include <linux/smp.h>
35 #include <linux/hrtimer.h>
37 #include <linux/slab.h>
38 #include <linux/export.h>
39 #include <asm/processor.h>
41 #include <asm/current.h>
42 #include <trace/events/kvm.h>
49 #define ioapic_debug(fmt,arg...) printk(KERN_WARNING fmt,##arg)
51 #define ioapic_debug(fmt, arg...)
53 static int ioapic_service(struct kvm_ioapic *vioapic, int irq,
56 static unsigned long ioapic_read_indirect(struct kvm_ioapic *ioapic,
60 unsigned long result = 0;
62 switch (ioapic->ioregsel) {
63 case IOAPIC_REG_VERSION:
64 result = ((((IOAPIC_NUM_PINS - 1) & 0xff) << 16)
65 | (IOAPIC_VERSION_ID & 0xff));
68 case IOAPIC_REG_APIC_ID:
69 case IOAPIC_REG_ARB_ID:
70 result = ((ioapic->id & 0xf) << 24);
75 u32 redir_index = (ioapic->ioregsel - 0x10) >> 1;
78 if (redir_index < IOAPIC_NUM_PINS)
80 ioapic->redirtbl[redir_index].bits;
82 redir_content = ~0ULL;
84 result = (ioapic->ioregsel & 0x1) ?
85 (redir_content >> 32) & 0xffffffff :
86 redir_content & 0xffffffff;
94 static void rtc_irq_eoi_tracking_reset(struct kvm_ioapic *ioapic)
96 ioapic->rtc_status.pending_eoi = 0;
97 bitmap_zero(ioapic->rtc_status.dest_map, KVM_MAX_VCPUS);
100 static void __rtc_irq_eoi_tracking_restore_one(struct kvm_vcpu *vcpu)
102 bool new_val, old_val;
103 struct kvm_ioapic *ioapic = vcpu->kvm->arch.vioapic;
104 union kvm_ioapic_redirect_entry *e;
106 e = &ioapic->redirtbl[RTC_GSI];
107 if (!kvm_apic_match_dest(vcpu, NULL, 0, e->fields.dest_id,
108 e->fields.dest_mode))
111 new_val = kvm_apic_pending_eoi(vcpu, e->fields.vector);
112 old_val = test_bit(vcpu->vcpu_id, ioapic->rtc_status.dest_map);
114 if (new_val == old_val)
118 __set_bit(vcpu->vcpu_id, ioapic->rtc_status.dest_map);
119 ioapic->rtc_status.pending_eoi++;
121 __clear_bit(vcpu->vcpu_id, ioapic->rtc_status.dest_map);
122 ioapic->rtc_status.pending_eoi--;
125 WARN_ON(ioapic->rtc_status.pending_eoi < 0);
128 void kvm_rtc_eoi_tracking_restore_one(struct kvm_vcpu *vcpu)
130 struct kvm_ioapic *ioapic = vcpu->kvm->arch.vioapic;
132 spin_lock(&ioapic->lock);
133 __rtc_irq_eoi_tracking_restore_one(vcpu);
134 spin_unlock(&ioapic->lock);
137 static void kvm_rtc_eoi_tracking_restore_all(struct kvm_ioapic *ioapic)
139 struct kvm_vcpu *vcpu;
142 if (RTC_GSI >= IOAPIC_NUM_PINS)
145 rtc_irq_eoi_tracking_reset(ioapic);
146 kvm_for_each_vcpu(i, vcpu, ioapic->kvm)
147 __rtc_irq_eoi_tracking_restore_one(vcpu);
150 static void rtc_irq_eoi(struct kvm_ioapic *ioapic, struct kvm_vcpu *vcpu)
152 if (test_and_clear_bit(vcpu->vcpu_id, ioapic->rtc_status.dest_map))
153 --ioapic->rtc_status.pending_eoi;
155 WARN_ON(ioapic->rtc_status.pending_eoi < 0);
158 static bool rtc_irq_check_coalesced(struct kvm_ioapic *ioapic)
160 if (ioapic->rtc_status.pending_eoi > 0)
161 return true; /* coalesced */
166 static int ioapic_set_irq(struct kvm_ioapic *ioapic, unsigned int irq,
167 int irq_level, bool line_status)
169 union kvm_ioapic_redirect_entry entry;
174 entry = ioapic->redirtbl[irq];
175 edge = (entry.fields.trig_mode == IOAPIC_EDGE_TRIG);
178 ioapic->irr &= ~mask;
184 * Return 0 for coalesced interrupts; for edge-triggered interrupts,
185 * this only happens if a previous edge has not been delivered due
186 * do masking. For level interrupts, the remote_irr field tells
187 * us if the interrupt is waiting for an EOI.
189 * RTC is special: it is edge-triggered, but userspace likes to know
190 * if it has been already ack-ed via EOI because coalesced RTC
191 * interrupts lead to time drift in Windows guests. So we track
192 * EOI manually for the RTC interrupt.
194 if (irq == RTC_GSI && line_status &&
195 rtc_irq_check_coalesced(ioapic)) {
200 old_irr = ioapic->irr;
202 if ((edge && old_irr == ioapic->irr) ||
203 (!edge && entry.fields.remote_irr)) {
208 ret = ioapic_service(ioapic, irq, line_status);
211 trace_kvm_ioapic_set_irq(entry.bits, irq, ret == 0);
215 static void kvm_ioapic_inject_all(struct kvm_ioapic *ioapic, unsigned long irr)
219 rtc_irq_eoi_tracking_reset(ioapic);
220 for_each_set_bit(idx, &irr, IOAPIC_NUM_PINS)
221 ioapic_set_irq(ioapic, idx, 1, true);
223 kvm_rtc_eoi_tracking_restore_all(ioapic);
227 static void update_handled_vectors(struct kvm_ioapic *ioapic)
229 DECLARE_BITMAP(handled_vectors, 256);
232 memset(handled_vectors, 0, sizeof(handled_vectors));
233 for (i = 0; i < IOAPIC_NUM_PINS; ++i)
234 __set_bit(ioapic->redirtbl[i].fields.vector, handled_vectors);
235 memcpy(ioapic->handled_vectors, handled_vectors,
236 sizeof(handled_vectors));
240 void kvm_ioapic_scan_entry(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap,
243 struct kvm_ioapic *ioapic = vcpu->kvm->arch.vioapic;
244 union kvm_ioapic_redirect_entry *e;
247 spin_lock(&ioapic->lock);
248 for (index = 0; index < IOAPIC_NUM_PINS; index++) {
249 e = &ioapic->redirtbl[index];
250 if (!e->fields.mask &&
251 (e->fields.trig_mode == IOAPIC_LEVEL_TRIG ||
252 kvm_irq_has_notifier(ioapic->kvm, KVM_IRQCHIP_IOAPIC,
253 index) || index == RTC_GSI)) {
254 if (kvm_apic_match_dest(vcpu, NULL, 0,
255 e->fields.dest_id, e->fields.dest_mode)) {
256 __set_bit(e->fields.vector,
257 (unsigned long *)eoi_exit_bitmap);
258 if (e->fields.trig_mode == IOAPIC_LEVEL_TRIG)
259 __set_bit(e->fields.vector,
260 (unsigned long *)tmr);
264 spin_unlock(&ioapic->lock);
268 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
270 struct kvm_ioapic *ioapic = kvm->arch.vioapic;
274 kvm_make_scan_ioapic_request(kvm);
277 void kvm_vcpu_request_scan_ioapic(struct kvm *kvm)
283 static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
286 bool mask_before, mask_after;
287 union kvm_ioapic_redirect_entry *e;
289 switch (ioapic->ioregsel) {
290 case IOAPIC_REG_VERSION:
291 /* Writes are ignored. */
294 case IOAPIC_REG_APIC_ID:
295 ioapic->id = (val >> 24) & 0xf;
298 case IOAPIC_REG_ARB_ID:
302 index = (ioapic->ioregsel - 0x10) >> 1;
304 ioapic_debug("change redir index %x val %x\n", index, val);
305 if (index >= IOAPIC_NUM_PINS)
307 e = &ioapic->redirtbl[index];
308 mask_before = e->fields.mask;
309 if (ioapic->ioregsel & 1) {
310 e->bits &= 0xffffffff;
311 e->bits |= (u64) val << 32;
313 e->bits &= ~0xffffffffULL;
314 e->bits |= (u32) val;
315 e->fields.remote_irr = 0;
317 update_handled_vectors(ioapic);
318 mask_after = e->fields.mask;
319 if (mask_before != mask_after)
320 kvm_fire_mask_notifiers(ioapic->kvm, KVM_IRQCHIP_IOAPIC, index, mask_after);
321 if (e->fields.trig_mode == IOAPIC_LEVEL_TRIG
322 && ioapic->irr & (1 << index))
323 ioapic_service(ioapic, index, false);
324 kvm_vcpu_request_scan_ioapic(ioapic->kvm);
329 static int ioapic_service(struct kvm_ioapic *ioapic, int irq, bool line_status)
331 union kvm_ioapic_redirect_entry *entry = &ioapic->redirtbl[irq];
332 struct kvm_lapic_irq irqe;
335 if (entry->fields.mask)
338 ioapic_debug("dest=%x dest_mode=%x delivery_mode=%x "
339 "vector=%x trig_mode=%x\n",
340 entry->fields.dest_id, entry->fields.dest_mode,
341 entry->fields.delivery_mode, entry->fields.vector,
342 entry->fields.trig_mode);
344 irqe.dest_id = entry->fields.dest_id;
345 irqe.vector = entry->fields.vector;
346 irqe.dest_mode = entry->fields.dest_mode;
347 irqe.trig_mode = entry->fields.trig_mode;
348 irqe.delivery_mode = entry->fields.delivery_mode << 8;
352 if (irqe.trig_mode == IOAPIC_EDGE_TRIG)
353 ioapic->irr &= ~(1 << irq);
355 if (irq == RTC_GSI && line_status) {
356 BUG_ON(ioapic->rtc_status.pending_eoi != 0);
357 ret = kvm_irq_delivery_to_apic(ioapic->kvm, NULL, &irqe,
358 ioapic->rtc_status.dest_map);
359 ioapic->rtc_status.pending_eoi = ret;
361 ret = kvm_irq_delivery_to_apic(ioapic->kvm, NULL, &irqe, NULL);
363 if (ret && irqe.trig_mode == IOAPIC_LEVEL_TRIG)
364 entry->fields.remote_irr = 1;
369 int kvm_ioapic_set_irq(struct kvm_ioapic *ioapic, int irq, int irq_source_id,
370 int level, bool line_status)
374 BUG_ON(irq < 0 || irq >= IOAPIC_NUM_PINS);
376 spin_lock(&ioapic->lock);
377 irq_level = __kvm_irq_line_state(&ioapic->irq_states[irq],
378 irq_source_id, level);
379 ret = ioapic_set_irq(ioapic, irq, irq_level, line_status);
381 spin_unlock(&ioapic->lock);
386 void kvm_ioapic_clear_all(struct kvm_ioapic *ioapic, int irq_source_id)
390 spin_lock(&ioapic->lock);
391 for (i = 0; i < KVM_IOAPIC_NUM_PINS; i++)
392 __clear_bit(irq_source_id, &ioapic->irq_states[i]);
393 spin_unlock(&ioapic->lock);
396 static void __kvm_ioapic_update_eoi(struct kvm_vcpu *vcpu,
397 struct kvm_ioapic *ioapic, int vector, int trigger_mode)
401 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
402 union kvm_ioapic_redirect_entry *ent = &ioapic->redirtbl[i];
404 if (ent->fields.vector != vector)
408 rtc_irq_eoi(ioapic, vcpu);
410 * We are dropping lock while calling ack notifiers because ack
411 * notifier callbacks for assigned devices call into IOAPIC
412 * recursively. Since remote_irr is cleared only after call
413 * to notifiers if the same vector will be delivered while lock
414 * is dropped it will be put into irr and will be delivered
415 * after ack notifier returns.
417 spin_unlock(&ioapic->lock);
418 kvm_notify_acked_irq(ioapic->kvm, KVM_IRQCHIP_IOAPIC, i);
419 spin_lock(&ioapic->lock);
421 if (trigger_mode != IOAPIC_LEVEL_TRIG)
424 ASSERT(ent->fields.trig_mode == IOAPIC_LEVEL_TRIG);
425 ent->fields.remote_irr = 0;
426 if (ioapic->irr & (1 << i))
427 ioapic_service(ioapic, i, false);
431 bool kvm_ioapic_handles_vector(struct kvm *kvm, int vector)
433 struct kvm_ioapic *ioapic = kvm->arch.vioapic;
435 return test_bit(vector, ioapic->handled_vectors);
438 void kvm_ioapic_update_eoi(struct kvm_vcpu *vcpu, int vector, int trigger_mode)
440 struct kvm_ioapic *ioapic = vcpu->kvm->arch.vioapic;
442 spin_lock(&ioapic->lock);
443 __kvm_ioapic_update_eoi(vcpu, ioapic, vector, trigger_mode);
444 spin_unlock(&ioapic->lock);
447 static inline struct kvm_ioapic *to_ioapic(struct kvm_io_device *dev)
449 return container_of(dev, struct kvm_ioapic, dev);
452 static inline int ioapic_in_range(struct kvm_ioapic *ioapic, gpa_t addr)
454 return ((addr >= ioapic->base_address &&
455 (addr < ioapic->base_address + IOAPIC_MEM_LENGTH)));
458 static int ioapic_mmio_read(struct kvm_io_device *this, gpa_t addr, int len,
461 struct kvm_ioapic *ioapic = to_ioapic(this);
463 if (!ioapic_in_range(ioapic, addr))
466 ioapic_debug("addr %lx\n", (unsigned long)addr);
467 ASSERT(!(addr & 0xf)); /* check alignment */
470 spin_lock(&ioapic->lock);
472 case IOAPIC_REG_SELECT:
473 result = ioapic->ioregsel;
476 case IOAPIC_REG_WINDOW:
477 result = ioapic_read_indirect(ioapic, addr, len);
484 spin_unlock(&ioapic->lock);
488 *(u64 *) val = result;
493 memcpy(val, (char *)&result, len);
496 printk(KERN_WARNING "ioapic: wrong length %d\n", len);
501 static int ioapic_mmio_write(struct kvm_io_device *this, gpa_t addr, int len,
504 struct kvm_ioapic *ioapic = to_ioapic(this);
506 if (!ioapic_in_range(ioapic, addr))
509 ioapic_debug("ioapic_mmio_write addr=%p len=%d val=%p\n",
510 (void*)addr, len, val);
511 ASSERT(!(addr & 0xf)); /* check alignment */
525 printk(KERN_WARNING "ioapic: Unsupported size %d\n", len);
530 spin_lock(&ioapic->lock);
532 case IOAPIC_REG_SELECT:
533 ioapic->ioregsel = data & 0xFF; /* 8-bit register */
536 case IOAPIC_REG_WINDOW:
537 ioapic_write_indirect(ioapic, data);
541 __kvm_ioapic_update_eoi(NULL, ioapic, data, IOAPIC_LEVEL_TRIG);
548 spin_unlock(&ioapic->lock);
552 static void kvm_ioapic_reset(struct kvm_ioapic *ioapic)
556 for (i = 0; i < IOAPIC_NUM_PINS; i++)
557 ioapic->redirtbl[i].fields.mask = 1;
558 ioapic->base_address = IOAPIC_DEFAULT_BASE_ADDRESS;
559 ioapic->ioregsel = 0;
562 rtc_irq_eoi_tracking_reset(ioapic);
563 update_handled_vectors(ioapic);
566 static const struct kvm_io_device_ops ioapic_mmio_ops = {
567 .read = ioapic_mmio_read,
568 .write = ioapic_mmio_write,
571 int kvm_ioapic_init(struct kvm *kvm)
573 struct kvm_ioapic *ioapic;
576 ioapic = kzalloc(sizeof(struct kvm_ioapic), GFP_KERNEL);
579 spin_lock_init(&ioapic->lock);
580 kvm->arch.vioapic = ioapic;
581 kvm_ioapic_reset(ioapic);
582 kvm_iodevice_init(&ioapic->dev, &ioapic_mmio_ops);
584 mutex_lock(&kvm->slots_lock);
585 ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, ioapic->base_address,
586 IOAPIC_MEM_LENGTH, &ioapic->dev);
587 mutex_unlock(&kvm->slots_lock);
589 kvm->arch.vioapic = NULL;
596 void kvm_ioapic_destroy(struct kvm *kvm)
598 struct kvm_ioapic *ioapic = kvm->arch.vioapic;
601 kvm_io_bus_unregister_dev(kvm, KVM_MMIO_BUS, &ioapic->dev);
602 kvm->arch.vioapic = NULL;
607 int kvm_get_ioapic(struct kvm *kvm, struct kvm_ioapic_state *state)
609 struct kvm_ioapic *ioapic = ioapic_irqchip(kvm);
613 spin_lock(&ioapic->lock);
614 memcpy(state, ioapic, sizeof(struct kvm_ioapic_state));
615 spin_unlock(&ioapic->lock);
619 int kvm_set_ioapic(struct kvm *kvm, struct kvm_ioapic_state *state)
621 struct kvm_ioapic *ioapic = ioapic_irqchip(kvm);
625 spin_lock(&ioapic->lock);
626 memcpy(ioapic, state, sizeof(struct kvm_ioapic_state));
628 update_handled_vectors(ioapic);
629 kvm_vcpu_request_scan_ioapic(kvm);
630 kvm_ioapic_inject_all(ioapic, state->irr);
631 spin_unlock(&ioapic->lock);