KVM: cleanup io_device code
[safe/jmp/linux-2.6] / arch / x86 / kvm / lapic.c
index 180ba73..4bfd458 100644 (file)
 #include <asm/current.h>
 #include <asm/apicdef.h>
 #include <asm/atomic.h>
+#include "kvm_cache_regs.h"
 #include "irq.h"
 
+#ifndef CONFIG_X86_64
+#define mod_64(x, y) ((x) - (y) * div64_u64(x, y))
+#else
+#define mod_64(x, y) ((x) % (y))
+#endif
+
 #define PRId64 "d"
 #define PRIx64 "llx"
 #define PRIu64 "u"
@@ -129,6 +136,11 @@ static inline int apic_lvtt_period(struct kvm_lapic *apic)
        return apic_get_reg(apic, APIC_LVTT) & APIC_LVT_TIMER_PERIODIC;
 }
 
+static inline int apic_lvt_nmi_mode(u32 lvt_val)
+{
+       return (lvt_val & (APIC_MODE_MASK | APIC_LVT_MASKED)) == APIC_DM_NMI;
+}
+
 static unsigned int apic_lvt_mask[APIC_LVT_NUM] = {
        LVT_MASK | APIC_LVT_TIMER_PERIODIC,     /* LVTT */
        LVT_MASK | APIC_MODE_MASK,      /* LVTTHMR */
@@ -184,20 +196,15 @@ int kvm_lapic_find_highest_irr(struct kvm_vcpu *vcpu)
 }
 EXPORT_SYMBOL_GPL(kvm_lapic_find_highest_irr);
 
-int kvm_apic_set_irq(struct kvm_vcpu *vcpu, u8 vec, u8 trig)
+static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
+                            int vector, int level, int trig_mode);
+
+int kvm_apic_set_irq(struct kvm_vcpu *vcpu, struct kvm_lapic_irq *irq)
 {
        struct kvm_lapic *apic = vcpu->arch.apic;
 
-       if (!apic_test_and_set_irr(vec, apic)) {
-               /* a new pending irq is set in IRR */
-               if (trig)
-                       apic_set_vector(vec, apic->regs + APIC_TMR);
-               else
-                       apic_clear_vector(vec, apic->regs + APIC_TMR);
-               kvm_vcpu_kick(apic->vcpu);
-               return 1;
-       }
-       return 0;
+       return __apic_accept_irq(apic, irq->delivery_mode, irq->vector,
+                       irq->level, irq->trig_mode);
 }
 
 static inline int apic_find_highest_isr(struct kvm_lapic *apic)
@@ -238,7 +245,7 @@ static void apic_set_tpr(struct kvm_lapic *apic, u32 tpr)
 
 int kvm_apic_match_physical_addr(struct kvm_lapic *apic, u16 dest)
 {
-       return kvm_apic_id(apic) == dest;
+       return dest == 0xff || kvm_apic_id(apic) == dest;
 }
 
 int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
@@ -267,37 +274,34 @@ int kvm_apic_match_logical_addr(struct kvm_lapic *apic, u8 mda)
        return result;
 }
 
-static int apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
+int kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
                           int short_hand, int dest, int dest_mode)
 {
        int result = 0;
        struct kvm_lapic *target = vcpu->arch.apic;
 
        apic_debug("target %p, source %p, dest 0x%x, "
-                  "dest_mode 0x%x, short_hand 0x%x",
+                  "dest_mode 0x%x, short_hand 0x%x\n",
                   target, source, dest, dest_mode, short_hand);
 
        ASSERT(!target);
        switch (short_hand) {
        case APIC_DEST_NOSHORT:
-               if (dest_mode == 0) {
+               if (dest_mode == 0)
                        /* Physical mode. */
-                       if ((dest == 0xFF) || (dest == kvm_apic_id(target)))
-                               result = 1;
-               } else
+                       result = kvm_apic_match_physical_addr(target, dest);
+               else
                        /* Logical mode. */
                        result = kvm_apic_match_logical_addr(target, dest);
                break;
        case APIC_DEST_SELF:
-               if (target == source)
-                       result = 1;
+               result = (target == source);
                break;
        case APIC_DEST_ALLINC:
                result = 1;
                break;
        case APIC_DEST_ALLBUT:
-               if (target != source)
-                       result = 1;
+               result = (target != source);
                break;
        default:
                printk(KERN_WARNING "Bad dest shorthand value %x\n",
@@ -315,20 +319,22 @@ static int apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
 static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
                             int vector, int level, int trig_mode)
 {
-       int orig_irr, result = 0;
+       int result = 0;
        struct kvm_vcpu *vcpu = apic->vcpu;
 
        switch (delivery_mode) {
-       case APIC_DM_FIXED:
        case APIC_DM_LOWEST:
+               vcpu->arch.apic_arb_prio++;
+       case APIC_DM_FIXED:
                /* FIXME add logic for vcpu on reset */
                if (unlikely(!apic_enabled(apic)))
                        break;
 
-               orig_irr = apic_test_and_set_irr(vector, apic);
-               if (orig_irr && trig_mode) {
-                       apic_debug("level trig mode repeatedly for vector %d",
-                                  vector);
+               result = !apic_test_and_set_irr(vector, apic);
+               if (!result) {
+                       if (trig_mode)
+                               apic_debug("level trig mode repeatedly for "
+                                               "vector %d", vector);
                        break;
                }
 
@@ -337,16 +343,7 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
                        apic_set_vector(vector, apic->regs + APIC_TMR);
                } else
                        apic_clear_vector(vector, apic->regs + APIC_TMR);
-
-               if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE)
-                       kvm_vcpu_kick(vcpu);
-               else if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED) {
-                       vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
-                       if (waitqueue_active(&vcpu->wq))
-                               wake_up_interruptible(&vcpu->wq);
-               }
-
-               result = (orig_irr == 0);
+               kvm_vcpu_kick(vcpu);
                break;
 
        case APIC_DM_REMRD:
@@ -358,11 +355,14 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
                break;
 
        case APIC_DM_NMI:
+               result = 1;
                kvm_inject_nmi(vcpu);
+               kvm_vcpu_kick(vcpu);
                break;
 
        case APIC_DM_INIT:
                if (level) {
+                       result = 1;
                        if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE)
                                printk(KERN_DEBUG
                                       "INIT on a runnable vcpu %d\n",
@@ -370,24 +370,30 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
                        vcpu->arch.mp_state = KVM_MP_STATE_INIT_RECEIVED;
                        kvm_vcpu_kick(vcpu);
                } else {
-                       printk(KERN_DEBUG
-                              "Ignoring de-assert INIT to vcpu %d\n",
-                              vcpu->vcpu_id);
+                       apic_debug("Ignoring de-assert INIT to vcpu %d\n",
+                                  vcpu->vcpu_id);
                }
-
                break;
 
        case APIC_DM_STARTUP:
-               printk(KERN_DEBUG "SIPI to vcpu %d vector 0x%02x\n",
-                      vcpu->vcpu_id, vector);
+               apic_debug("SIPI to vcpu %d vector 0x%02x\n",
+                          vcpu->vcpu_id, vector);
                if (vcpu->arch.mp_state == KVM_MP_STATE_INIT_RECEIVED) {
+                       result = 1;
                        vcpu->arch.sipi_vector = vector;
                        vcpu->arch.mp_state = KVM_MP_STATE_SIPI_RECEIVED;
-                       if (waitqueue_active(&vcpu->wq))
-                               wake_up_interruptible(&vcpu->wq);
+                       kvm_vcpu_kick(vcpu);
                }
                break;
 
+       case APIC_DM_EXTINT:
+               /*
+                * Should only be called by kvm_apic_local_deliver() with LVT0,
+                * before NMI watchdog was enabled. Already handled by
+                * kvm_apic_accept_pic_intr().
+                */
+               break;
+
        default:
                printk(KERN_ERR "TODO: unsupported delivery mode %x\n",
                       delivery_mode);
@@ -396,49 +402,15 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
        return result;
 }
 
-static struct kvm_lapic *kvm_apic_round_robin(struct kvm *kvm, u8 vector,
-                                      unsigned long bitmap)
+int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2)
 {
-       int last;
-       int next;
-       struct kvm_lapic *apic = NULL;
-
-       last = kvm->arch.round_robin_prev_vcpu;
-       next = last;
-
-       do {
-               if (++next == KVM_MAX_VCPUS)
-                       next = 0;
-               if (kvm->vcpus[next] == NULL || !test_bit(next, &bitmap))
-                       continue;
-               apic = kvm->vcpus[next]->arch.apic;
-               if (apic && apic_enabled(apic))
-                       break;
-               apic = NULL;
-       } while (next != last);
-       kvm->arch.round_robin_prev_vcpu = next;
-
-       if (!apic)
-               printk(KERN_DEBUG "vcpu not ready for apic_round_robin\n");
-
-       return apic;
-}
-
-struct kvm_vcpu *kvm_get_lowest_prio_vcpu(struct kvm *kvm, u8 vector,
-               unsigned long bitmap)
-{
-       struct kvm_lapic *apic;
-
-       apic = kvm_apic_round_robin(kvm, vector, bitmap);
-       if (apic)
-               return apic->vcpu;
-       return NULL;
+       return vcpu1->arch.apic_arb_prio - vcpu2->arch.apic_arb_prio;
 }
 
 static void apic_set_eoi(struct kvm_lapic *apic)
 {
        int vector = apic_find_highest_isr(apic);
-
+       int trigger_mode;
        /*
         * Not every write EOI will has corresponding ISR,
         * one example is when Kernel check timer on setup_IO_APIC
@@ -450,104 +422,55 @@ static void apic_set_eoi(struct kvm_lapic *apic)
        apic_update_ppr(apic);
 
        if (apic_test_and_clear_vector(vector, apic->regs + APIC_TMR))
-               kvm_ioapic_update_eoi(apic->vcpu->kvm, vector);
+               trigger_mode = IOAPIC_LEVEL_TRIG;
+       else
+               trigger_mode = IOAPIC_EDGE_TRIG;
+       kvm_ioapic_update_eoi(apic->vcpu->kvm, vector, trigger_mode);
 }
 
 static void apic_send_ipi(struct kvm_lapic *apic)
 {
        u32 icr_low = apic_get_reg(apic, APIC_ICR);
        u32 icr_high = apic_get_reg(apic, APIC_ICR2);
+       struct kvm_lapic_irq irq;
 
-       unsigned int dest = GET_APIC_DEST_FIELD(icr_high);
-       unsigned int short_hand = icr_low & APIC_SHORT_MASK;
-       unsigned int trig_mode = icr_low & APIC_INT_LEVELTRIG;
-       unsigned int level = icr_low & APIC_INT_ASSERT;
-       unsigned int dest_mode = icr_low & APIC_DEST_MASK;
-       unsigned int delivery_mode = icr_low & APIC_MODE_MASK;
-       unsigned int vector = icr_low & APIC_VECTOR_MASK;
-
-       struct kvm_vcpu *target;
-       struct kvm_vcpu *vcpu;
-       unsigned long lpr_map = 0;
-       int i;
+       irq.vector = icr_low & APIC_VECTOR_MASK;
+       irq.delivery_mode = icr_low & APIC_MODE_MASK;
+       irq.dest_mode = icr_low & APIC_DEST_MASK;
+       irq.level = icr_low & APIC_INT_ASSERT;
+       irq.trig_mode = icr_low & APIC_INT_LEVELTRIG;
+       irq.shorthand = icr_low & APIC_SHORT_MASK;
+       irq.dest_id = GET_APIC_DEST_FIELD(icr_high);
 
        apic_debug("icr_high 0x%x, icr_low 0x%x, "
                   "short_hand 0x%x, dest 0x%x, trig_mode 0x%x, level 0x%x, "
                   "dest_mode 0x%x, delivery_mode 0x%x, vector 0x%x\n",
-                  icr_high, icr_low, short_hand, dest,
-                  trig_mode, level, dest_mode, delivery_mode, vector);
-
-       for (i = 0; i < KVM_MAX_VCPUS; i++) {
-               vcpu = apic->vcpu->kvm->vcpus[i];
-               if (!vcpu)
-                       continue;
-
-               if (vcpu->arch.apic &&
-                   apic_match_dest(vcpu, apic, short_hand, dest, dest_mode)) {
-                       if (delivery_mode == APIC_DM_LOWEST)
-                               set_bit(vcpu->vcpu_id, &lpr_map);
-                       else
-                               __apic_accept_irq(vcpu->arch.apic, delivery_mode,
-                                                 vector, level, trig_mode);
-               }
-       }
+                  icr_high, icr_low, irq.shorthand, irq.dest_id,
+                  irq.trig_mode, irq.level, irq.dest_mode, irq.delivery_mode,
+                  irq.vector);
 
-       if (delivery_mode == APIC_DM_LOWEST) {
-               target = kvm_get_lowest_prio_vcpu(vcpu->kvm, vector, lpr_map);
-               if (target != NULL)
-                       __apic_accept_irq(target->arch.apic, delivery_mode,
-                                         vector, level, trig_mode);
-       }
+       kvm_irq_delivery_to_apic(apic->vcpu->kvm, apic, &irq);
 }
 
 static u32 apic_get_tmcct(struct kvm_lapic *apic)
 {
-       u64 counter_passed;
-       ktime_t passed, now;
+       ktime_t remaining;
+       s64 ns;
        u32 tmcct;
 
        ASSERT(apic != NULL);
 
-       now = apic->timer.dev.base->get_time();
-       tmcct = apic_get_reg(apic, APIC_TMICT);
-
        /* if initial count is 0, current count should also be 0 */
-       if (tmcct == 0)
+       if (apic_get_reg(apic, APIC_TMICT) == 0)
                return 0;
 
-       if (unlikely(ktime_to_ns(now) <=
-               ktime_to_ns(apic->timer.last_update))) {
-               /* Wrap around */
-               passed = ktime_add(( {
-                                   (ktime_t) {
-                                   .tv64 = KTIME_MAX -
-                                   (apic->timer.last_update).tv64}; }
-                                  ), now);
-               apic_debug("time elapsed\n");
-       } else
-               passed = ktime_sub(now, apic->timer.last_update);
-
-       counter_passed = div64_u64(ktime_to_ns(passed),
-                                  (APIC_BUS_CYCLE_NS * apic->timer.divide_count));
-
-       if (counter_passed > tmcct) {
-               if (unlikely(!apic_lvtt_period(apic))) {
-                       /* one-shot timers stick at 0 until reset */
-                       tmcct = 0;
-               } else {
-                       /*
-                        * periodic timers reset to APIC_TMICT when they
-                        * hit 0. The while loop simulates this happening N
-                        * times. (counter_passed %= tmcct) would also work,
-                        * but might be slower or not work on 32-bit??
-                        */
-                       while (counter_passed > tmcct)
-                               counter_passed -= tmcct;
-                       tmcct -= counter_passed;
-               }
-       } else {
-               tmcct -= counter_passed;
-       }
+       remaining = hrtimer_expires_remaining(&apic->lapic_timer.timer);
+       if (ktime_to_ns(remaining) < 0)
+               remaining = ktime_set(0, 0);
+
+       ns = mod_64(ktime_to_ns(remaining), apic->lapic_timer.period);
+       tmcct = div64_u64(ns,
+                        (APIC_BUS_CYCLE_NS * apic->divide_count));
 
        return tmcct;
 }
@@ -558,8 +481,7 @@ static void __report_tpr_access(struct kvm_lapic *apic, bool write)
        struct kvm_run *run = vcpu->run;
 
        set_bit(KVM_REQ_REPORT_TPR_ACCESS, &vcpu->requests);
-       kvm_x86_ops->cache_regs(vcpu);
-       run->tpr_access.rip = vcpu->arch.rip;
+       run->tpr_access.rip = kvm_rip_read(vcpu);
        run->tpr_access.is_write = write;
 }
 
@@ -600,10 +522,15 @@ static u32 __apic_read(struct kvm_lapic *apic, unsigned int offset)
        return val;
 }
 
+static inline struct kvm_lapic *to_lapic(struct kvm_io_device *dev)
+{
+       return container_of(dev, struct kvm_lapic, dev);
+}
+
 static void apic_mmio_read(struct kvm_io_device *this,
                           gpa_t address, int len, void *data)
 {
-       struct kvm_lapic *apic = (struct kvm_lapic *)this->private;
+       struct kvm_lapic *apic = to_lapic(this);
        unsigned int offset = address - apic->base_address;
        unsigned char alignment = offset & 0xf;
        u32 result;
@@ -635,27 +562,25 @@ static void update_divide_count(struct kvm_lapic *apic)
        tdcr = apic_get_reg(apic, APIC_TDCR);
        tmp1 = tdcr & 0xf;
        tmp2 = ((tmp1 & 0x3) | ((tmp1 & 0x8) >> 1)) + 1;
-       apic->timer.divide_count = 0x1 << (tmp2 & 0x7);
+       apic->divide_count = 0x1 << (tmp2 & 0x7);
 
        apic_debug("timer divide count is 0x%x\n",
-                                  apic->timer.divide_count);
+                                  apic->divide_count);
 }
 
 static void start_apic_timer(struct kvm_lapic *apic)
 {
-       ktime_t now = apic->timer.dev.base->get_time();
-
-       apic->timer.last_update = now;
+       ktime_t now = apic->lapic_timer.timer.base->get_time();
 
-       apic->timer.period = apic_get_reg(apic, APIC_TMICT) *
-                   APIC_BUS_CYCLE_NS * apic->timer.divide_count;
-       atomic_set(&apic->timer.pending, 0);
+       apic->lapic_timer.period = apic_get_reg(apic, APIC_TMICT) *
+                   APIC_BUS_CYCLE_NS * apic->divide_count;
+       atomic_set(&apic->lapic_timer.pending, 0);
 
-       if (!apic->timer.period)
+       if (!apic->lapic_timer.period)
                return;
 
-       hrtimer_start(&apic->timer.dev,
-                     ktime_add_ns(now, apic->timer.period),
+       hrtimer_start(&apic->lapic_timer.timer,
+                     ktime_add_ns(now, apic->lapic_timer.period),
                      HRTIMER_MODE_ABS);
 
        apic_debug("%s: bus cycle is %" PRId64 "ns, now 0x%016"
@@ -664,15 +589,29 @@ static void start_apic_timer(struct kvm_lapic *apic)
                           "expire @ 0x%016" PRIx64 ".\n", __func__,
                           APIC_BUS_CYCLE_NS, ktime_to_ns(now),
                           apic_get_reg(apic, APIC_TMICT),
-                          apic->timer.period,
+                          apic->lapic_timer.period,
                           ktime_to_ns(ktime_add_ns(now,
-                                       apic->timer.period)));
+                                       apic->lapic_timer.period)));
+}
+
+static void apic_manage_nmi_watchdog(struct kvm_lapic *apic, u32 lvt0_val)
+{
+       int nmi_wd_enabled = apic_lvt_nmi_mode(apic_get_reg(apic, APIC_LVT0));
+
+       if (apic_lvt_nmi_mode(lvt0_val)) {
+               if (!nmi_wd_enabled) {
+                       apic_debug("Receive NMI setting on APIC_LVT0 "
+                                  "for cpu %d\n", apic->vcpu->vcpu_id);
+                       apic->vcpu->kvm->arch.vapics_in_nmi_mode++;
+               }
+       } else if (nmi_wd_enabled)
+               apic->vcpu->kvm->arch.vapics_in_nmi_mode--;
 }
 
 static void apic_mmio_write(struct kvm_io_device *this,
                            gpa_t address, int len, const void *data)
 {
-       struct kvm_lapic *apic = (struct kvm_lapic *)this->private;
+       struct kvm_lapic *apic = to_lapic(this);
        unsigned int offset = address - apic->base_address;
        unsigned char alignment = offset & 0xf;
        u32 val;
@@ -683,9 +622,9 @@ static void apic_mmio_write(struct kvm_io_device *this,
         * Refer SDM 8.4.1
         */
        if (len != 4 || alignment) {
-               if (printk_ratelimit())
-                       printk(KERN_ERR "apic write: bad size=%d %lx\n",
-                              len, (long)address);
+               /* Don't shout loud, $infamous_os would cause only noise. */
+               apic_debug("apic write: bad size=%d %lx\n",
+                          len, (long)address);
                return;
        }
 
@@ -734,7 +673,7 @@ static void apic_mmio_write(struct kvm_io_device *this,
                                apic_set_reg(apic, APIC_LVTT + 0x10 * i,
                                             lvt_val | APIC_LVT_MASKED);
                        }
-                       atomic_set(&apic->timer.pending, 0);
+                       atomic_set(&apic->lapic_timer.pending, 0);
 
                }
                break;
@@ -749,10 +688,11 @@ static void apic_mmio_write(struct kvm_io_device *this,
                apic_set_reg(apic, APIC_ICR2, val & 0xff000000);
                break;
 
+       case APIC_LVT0:
+               apic_manage_nmi_watchdog(apic, val);
        case APIC_LVTT:
        case APIC_LVTTHMR:
        case APIC_LVTPC:
-       case APIC_LVT0:
        case APIC_LVT1:
        case APIC_LVTERR:
                /* TODO: Check vector */
@@ -765,7 +705,7 @@ static void apic_mmio_write(struct kvm_io_device *this,
                break;
 
        case APIC_TMICT:
-               hrtimer_cancel(&apic->timer.dev);
+               hrtimer_cancel(&apic->lapic_timer.timer);
                apic_set_reg(apic, APIC_TMICT, val);
                start_apic_timer(apic);
                return;
@@ -788,7 +728,7 @@ static void apic_mmio_write(struct kvm_io_device *this,
 static int apic_mmio_range(struct kvm_io_device *this, gpa_t addr,
                           int len, int size)
 {
-       struct kvm_lapic *apic = (struct kvm_lapic *)this->private;
+       struct kvm_lapic *apic = to_lapic(this);
        int ret = 0;
 
 
@@ -805,7 +745,7 @@ void kvm_free_lapic(struct kvm_vcpu *vcpu)
        if (!vcpu->arch.apic)
                return;
 
-       hrtimer_cancel(&vcpu->arch.apic->timer.dev);
+       hrtimer_cancel(&vcpu->arch.apic->lapic_timer.timer);
 
        if (vcpu->arch.apic->regs_page)
                __free_page(vcpu->arch.apic->regs_page);
@@ -883,7 +823,7 @@ void kvm_lapic_reset(struct kvm_vcpu *vcpu)
        ASSERT(apic != NULL);
 
        /* Stop the timer in case it's a reset to an active apic */
-       hrtimer_cancel(&apic->timer.dev);
+       hrtimer_cancel(&apic->lapic_timer.timer);
 
        apic_set_reg(apic, APIC_ID, vcpu->vcpu_id << 24);
        apic_set_reg(apic, APIC_LVR, APIC_VERSION);
@@ -908,11 +848,13 @@ void kvm_lapic_reset(struct kvm_vcpu *vcpu)
                apic_set_reg(apic, APIC_TMR + 0x10 * i, 0);
        }
        update_divide_count(apic);
-       atomic_set(&apic->timer.pending, 0);
+       atomic_set(&apic->lapic_timer.pending, 0);
        if (vcpu->vcpu_id == 0)
                vcpu->arch.apic_base |= MSR_IA32_APICBASE_BSP;
        apic_update_ppr(apic);
 
+       vcpu->arch.apic_arb_prio = 0;
+
        apic_debug(KERN_INFO "%s: vcpu=%p, id=%d, base_msr="
                   "0x%016" PRIx64 ", base_address=0x%0lx.\n", __func__,
                   vcpu, kvm_apic_id(apic),
@@ -920,16 +862,14 @@ void kvm_lapic_reset(struct kvm_vcpu *vcpu)
 }
 EXPORT_SYMBOL_GPL(kvm_lapic_reset);
 
-int kvm_lapic_enabled(struct kvm_vcpu *vcpu)
+bool kvm_apic_present(struct kvm_vcpu *vcpu)
 {
-       struct kvm_lapic *apic = vcpu->arch.apic;
-       int ret = 0;
-
-       if (!apic)
-               return 0;
-       ret = apic_enabled(apic);
+       return vcpu->arch.apic && apic_hw_enabled(vcpu->arch.apic);
+}
 
-       return ret;
+int kvm_lapic_enabled(struct kvm_vcpu *vcpu)
+{
+       return kvm_apic_present(vcpu) && apic_sw_enabled(vcpu->arch.apic);
 }
 EXPORT_SYMBOL_GPL(kvm_lapic_enabled);
 
@@ -939,25 +879,11 @@ EXPORT_SYMBOL_GPL(kvm_lapic_enabled);
  *----------------------------------------------------------------------
  */
 
-/* TODO: make sure __apic_timer_fn runs in current pCPU */
-static int __apic_timer_fn(struct kvm_lapic *apic)
+static bool lapic_is_periodic(struct kvm_timer *ktimer)
 {
-       int result = 0;
-       wait_queue_head_t *q = &apic->vcpu->wq;
-
-       atomic_inc(&apic->timer.pending);
-       set_bit(KVM_REQ_PENDING_TIMER, &apic->vcpu->requests);
-       if (waitqueue_active(q)) {
-               apic->vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
-               wake_up_interruptible(q);
-       }
-       if (apic_lvtt_period(apic)) {
-               result = 1;
-               apic->timer.dev.expires = ktime_add_ns(
-                                       apic->timer.dev.expires,
-                                       apic->timer.period);
-       }
-       return result;
+       struct kvm_lapic *apic = container_of(ktimer, struct kvm_lapic,
+                                             lapic_timer);
+       return apic_lvtt_period(apic);
 }
 
 int apic_has_pending_timer(struct kvm_vcpu *vcpu)
@@ -965,33 +891,42 @@ int apic_has_pending_timer(struct kvm_vcpu *vcpu)
        struct kvm_lapic *lapic = vcpu->arch.apic;
 
        if (lapic && apic_enabled(lapic) && apic_lvt_enabled(lapic, APIC_LVTT))
-               return atomic_read(&lapic->timer.pending);
+               return atomic_read(&lapic->lapic_timer.pending);
 
        return 0;
 }
 
-static int __inject_apic_timer_irq(struct kvm_lapic *apic)
+static int kvm_apic_local_deliver(struct kvm_lapic *apic, int lvt_type)
 {
-       int vector;
-
-       vector = apic_lvt_vector(apic, APIC_LVTT);
-       return __apic_accept_irq(apic, APIC_DM_FIXED, vector, 1, 0);
+       u32 reg = apic_get_reg(apic, lvt_type);
+       int vector, mode, trig_mode;
+
+       if (apic_hw_enabled(apic) && !(reg & APIC_LVT_MASKED)) {
+               vector = reg & APIC_VECTOR_MASK;
+               mode = reg & APIC_MODE_MASK;
+               trig_mode = reg & APIC_LVT_LEVEL_TRIGGER;
+               return __apic_accept_irq(apic, mode, vector, 1, trig_mode);
+       }
+       return 0;
 }
 
-static enum hrtimer_restart apic_timer_fn(struct hrtimer *data)
+void kvm_apic_nmi_wd_deliver(struct kvm_vcpu *vcpu)
 {
-       struct kvm_lapic *apic;
-       int restart_timer = 0;
+       struct kvm_lapic *apic = vcpu->arch.apic;
 
-       apic = container_of(data, struct kvm_lapic, timer.dev);
+       if (apic)
+               kvm_apic_local_deliver(apic, APIC_LVT0);
+}
 
-       restart_timer = __apic_timer_fn(apic);
+static struct kvm_timer_ops lapic_timer_ops = {
+       .is_periodic = lapic_is_periodic,
+};
 
-       if (restart_timer)
-               return HRTIMER_RESTART;
-       else
-               return HRTIMER_NORESTART;
-}
+static const struct kvm_io_device_ops apic_mmio_ops = {
+       .read     = apic_mmio_read,
+       .write    = apic_mmio_write,
+       .in_range = apic_mmio_range,
+};
 
 int kvm_create_lapic(struct kvm_vcpu *vcpu)
 {
@@ -1016,16 +951,18 @@ int kvm_create_lapic(struct kvm_vcpu *vcpu)
        memset(apic->regs, 0, PAGE_SIZE);
        apic->vcpu = vcpu;
 
-       hrtimer_init(&apic->timer.dev, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
-       apic->timer.dev.function = apic_timer_fn;
+       hrtimer_init(&apic->lapic_timer.timer, CLOCK_MONOTONIC,
+                    HRTIMER_MODE_ABS);
+       apic->lapic_timer.timer.function = kvm_timer_fn;
+       apic->lapic_timer.t_ops = &lapic_timer_ops;
+       apic->lapic_timer.kvm = vcpu->kvm;
+       apic->lapic_timer.vcpu_id = vcpu->vcpu_id;
+
        apic->base_address = APIC_DEFAULT_PHYS_BASE;
        vcpu->arch.apic_base = APIC_DEFAULT_PHYS_BASE;
 
        kvm_lapic_reset(vcpu);
-       apic->dev.read = apic_mmio_read;
-       apic->dev.write = apic_mmio_write;
-       apic->dev.in_range = apic_mmio_range;
-       apic->dev.private = apic;
+       kvm_iodevice_init(&apic->dev, &apic_mmio_ops);
 
        return 0;
 nomem_free_apic:
@@ -1070,23 +1007,12 @@ void kvm_inject_apic_timer_irqs(struct kvm_vcpu *vcpu)
 {
        struct kvm_lapic *apic = vcpu->arch.apic;
 
-       if (apic && apic_lvt_enabled(apic, APIC_LVTT) &&
-               atomic_read(&apic->timer.pending) > 0) {
-               if (__inject_apic_timer_irq(apic))
-                       atomic_dec(&apic->timer.pending);
+       if (apic && atomic_read(&apic->lapic_timer.pending) > 0) {
+               if (kvm_apic_local_deliver(apic, APIC_LVTT))
+                       atomic_dec(&apic->lapic_timer.pending);
        }
 }
 
-void kvm_apic_timer_intr_post(struct kvm_vcpu *vcpu, int vec)
-{
-       struct kvm_lapic *apic = vcpu->arch.apic;
-
-       if (apic && apic_lvt_vector(apic, APIC_LVTT) == vec)
-               apic->timer.last_update = ktime_add_ns(
-                               apic->timer.last_update,
-                               apic->timer.period);
-}
-
 int kvm_get_apic_interrupt(struct kvm_vcpu *vcpu)
 {
        int vector = kvm_apic_has_interrupt(vcpu);
@@ -1109,7 +1035,7 @@ void kvm_apic_post_state_restore(struct kvm_vcpu *vcpu)
                             MSR_IA32_APICBASE_BASE;
        apic_set_reg(apic, APIC_LVR, APIC_VERSION);
        apic_update_ppr(apic);
-       hrtimer_cancel(&apic->timer.dev);
+       hrtimer_cancel(&apic->lapic_timer.timer);
        update_divide_count(apic);
        start_apic_timer(apic);
 }
@@ -1122,9 +1048,9 @@ void __kvm_migrate_apic_timer(struct kvm_vcpu *vcpu)
        if (!apic)
                return;
 
-       timer = &apic->timer.dev;
+       timer = &apic->lapic_timer.timer;
        if (hrtimer_cancel(timer))
-               hrtimer_start(timer, timer->expires, HRTIMER_MODE_ABS);
+               hrtimer_start_expires(timer, HRTIMER_MODE_ABS);
 }
 
 void kvm_lapic_sync_from_vapic(struct kvm_vcpu *vcpu)