KVM: In kernel PIT model
authorSheng Yang <sheng.yang@intel.com>
Sun, 27 Jan 2008 21:10:22 +0000 (05:10 +0800)
committerAvi Kivity <avi@qumranet.com>
Sun, 27 Apr 2008 09:00:21 +0000 (12:00 +0300)
The patch moves the PIT model from userspace to kernel, and increases
the timer accuracy greatly.

[marcelo: make last_injected_time per-guest]

Signed-off-by: Sheng Yang <sheng.yang@intel.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Tested-and-Acked-by: Alex Davis <alex14641@yahoo.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
arch/x86/kvm/Makefile
arch/x86/kvm/i8254.c [new file with mode: 0644]
arch/x86/kvm/i8254.h [new file with mode: 0644]
arch/x86/kvm/irq.c
arch/x86/kvm/x86.c
include/asm-x86/kvm_host.h
include/linux/kvm.h

index ffdd0b3..4d0c22e 100644 (file)
@@ -6,7 +6,8 @@ common-objs = $(addprefix ../../../virt/kvm/, kvm_main.o ioapic.o)
 
 EXTRA_CFLAGS += -Ivirt/kvm -Iarch/x86/kvm
 
-kvm-objs := $(common-objs) x86.o mmu.o x86_emulate.o i8259.o irq.o lapic.o
+kvm-objs := $(common-objs) x86.o mmu.o x86_emulate.o i8259.o irq.o lapic.o \
+       i8254.o
 obj-$(CONFIG_KVM) += kvm.o
 kvm-intel-objs = vmx.o
 obj-$(CONFIG_KVM_INTEL) += kvm-intel.o
diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
new file mode 100644 (file)
index 0000000..c743509
--- /dev/null
@@ -0,0 +1,586 @@
+/*
+ * 8253/8254 interval timer emulation
+ *
+ * Copyright (c) 2003-2004 Fabrice Bellard
+ * Copyright (c) 2006 Intel Corporation
+ * Copyright (c) 2007 Keir Fraser, XenSource Inc
+ * Copyright (c) 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ *
+ * Authors:
+ *   Sheng Yang <sheng.yang@intel.com>
+ *   Based on QEMU and Xen.
+ */
+
+#include <linux/kvm_host.h>
+
+#include "irq.h"
+#include "i8254.h"
+
+#ifndef CONFIG_X86_64
+#define mod_64(x, y) ((x) - (y) * div64_64(x, y))
+#else
+#define mod_64(x, y) ((x) % (y))
+#endif
+
+#define RW_STATE_LSB 1
+#define RW_STATE_MSB 2
+#define RW_STATE_WORD0 3
+#define RW_STATE_WORD1 4
+
+/* Compute with 96 bit intermediate result: (a*b)/c */
+static u64 muldiv64(u64 a, u32 b, u32 c)
+{
+       union {
+               u64 ll;
+               struct {
+                       u32 low, high;
+               } l;
+       } u, res;
+       u64 rl, rh;
+
+       u.ll = a;
+       rl = (u64)u.l.low * (u64)b;
+       rh = (u64)u.l.high * (u64)b;
+       rh += (rl >> 32);
+       res.l.high = div64_64(rh, c);
+       res.l.low = div64_64(((mod_64(rh, c) << 32) + (rl & 0xffffffff)), c);
+       return res.ll;
+}
+
+static void pit_set_gate(struct kvm *kvm, int channel, u32 val)
+{
+       struct kvm_kpit_channel_state *c =
+               &kvm->arch.vpit->pit_state.channels[channel];
+
+       WARN_ON(!mutex_is_locked(&kvm->arch.vpit->pit_state.lock));
+
+       switch (c->mode) {
+       default:
+       case 0:
+       case 4:
+               /* XXX: just disable/enable counting */
+               break;
+       case 1:
+       case 2:
+       case 3:
+       case 5:
+               /* Restart counting on rising edge. */
+               if (c->gate < val)
+                       c->count_load_time = ktime_get();
+               break;
+       }
+
+       c->gate = val;
+}
+
+int pit_get_gate(struct kvm *kvm, int channel)
+{
+       WARN_ON(!mutex_is_locked(&kvm->arch.vpit->pit_state.lock));
+
+       return kvm->arch.vpit->pit_state.channels[channel].gate;
+}
+
+static int pit_get_count(struct kvm *kvm, int channel)
+{
+       struct kvm_kpit_channel_state *c =
+               &kvm->arch.vpit->pit_state.channels[channel];
+       s64 d, t;
+       int counter;
+
+       WARN_ON(!mutex_is_locked(&kvm->arch.vpit->pit_state.lock));
+
+       t = ktime_to_ns(ktime_sub(ktime_get(), c->count_load_time));
+       d = muldiv64(t, KVM_PIT_FREQ, NSEC_PER_SEC);
+
+       switch (c->mode) {
+       case 0:
+       case 1:
+       case 4:
+       case 5:
+               counter = (c->count - d) & 0xffff;
+               break;
+       case 3:
+               /* XXX: may be incorrect for odd counts */
+               counter = c->count - (mod_64((2 * d), c->count));
+               break;
+       default:
+               counter = c->count - mod_64(d, c->count);
+               break;
+       }
+       return counter;
+}
+
+static int pit_get_out(struct kvm *kvm, int channel)
+{
+       struct kvm_kpit_channel_state *c =
+               &kvm->arch.vpit->pit_state.channels[channel];
+       s64 d, t;
+       int out;
+
+       WARN_ON(!mutex_is_locked(&kvm->arch.vpit->pit_state.lock));
+
+       t = ktime_to_ns(ktime_sub(ktime_get(), c->count_load_time));
+       d = muldiv64(t, KVM_PIT_FREQ, NSEC_PER_SEC);
+
+       switch (c->mode) {
+       default:
+       case 0:
+               out = (d >= c->count);
+               break;
+       case 1:
+               out = (d < c->count);
+               break;
+       case 2:
+               out = ((mod_64(d, c->count) == 0) && (d != 0));
+               break;
+       case 3:
+               out = (mod_64(d, c->count) < ((c->count + 1) >> 1));
+               break;
+       case 4:
+       case 5:
+               out = (d == c->count);
+               break;
+       }
+
+       return out;
+}
+
+static void pit_latch_count(struct kvm *kvm, int channel)
+{
+       struct kvm_kpit_channel_state *c =
+               &kvm->arch.vpit->pit_state.channels[channel];
+
+       WARN_ON(!mutex_is_locked(&kvm->arch.vpit->pit_state.lock));
+
+       if (!c->count_latched) {
+               c->latched_count = pit_get_count(kvm, channel);
+               c->count_latched = c->rw_mode;
+       }
+}
+
+static void pit_latch_status(struct kvm *kvm, int channel)
+{
+       struct kvm_kpit_channel_state *c =
+               &kvm->arch.vpit->pit_state.channels[channel];
+
+       WARN_ON(!mutex_is_locked(&kvm->arch.vpit->pit_state.lock));
+
+       if (!c->status_latched) {
+               /* TODO: Return NULL COUNT (bit 6). */
+               c->status = ((pit_get_out(kvm, channel) << 7) |
+                               (c->rw_mode << 4) |
+                               (c->mode << 1) |
+                               c->bcd);
+               c->status_latched = 1;
+       }
+}
+
+int __pit_timer_fn(struct kvm_kpit_state *ps)
+{
+       struct kvm_vcpu *vcpu0 = ps->pit->kvm->vcpus[0];
+       struct kvm_kpit_timer *pt = &ps->pit_timer;
+
+       atomic_inc(&pt->pending);
+       smp_mb__after_atomic_inc();
+       /* FIXME: handle case where the guest is in guest mode */
+       if (vcpu0 && waitqueue_active(&vcpu0->wq)) {
+               vcpu0->arch.mp_state = VCPU_MP_STATE_RUNNABLE;
+               wake_up_interruptible(&vcpu0->wq);
+       }
+
+       pt->timer.expires = ktime_add_ns(pt->timer.expires, pt->period);
+       pt->scheduled = ktime_to_ns(pt->timer.expires);
+
+       return (pt->period == 0 ? 0 : 1);
+}
+
+static enum hrtimer_restart pit_timer_fn(struct hrtimer *data)
+{
+       struct kvm_kpit_state *ps;
+       int restart_timer = 0;
+
+       ps = container_of(data, struct kvm_kpit_state, pit_timer.timer);
+
+       restart_timer = __pit_timer_fn(ps);
+
+       if (restart_timer)
+               return HRTIMER_RESTART;
+       else
+               return HRTIMER_NORESTART;
+}
+
+static void destroy_pit_timer(struct kvm_kpit_timer *pt)
+{
+       pr_debug("pit: execute del timer!\n");
+       hrtimer_cancel(&pt->timer);
+}
+
+static void create_pit_timer(struct kvm_kpit_timer *pt, u32 val, int is_period)
+{
+       s64 interval;
+
+       interval = muldiv64(val, NSEC_PER_SEC, KVM_PIT_FREQ);
+
+       pr_debug("pit: create pit timer, interval is %llu nsec\n", interval);
+
+       /* TODO The new value only affected after the retriggered */
+       hrtimer_cancel(&pt->timer);
+       pt->period = (is_period == 0) ? 0 : interval;
+       pt->timer.function = pit_timer_fn;
+       atomic_set(&pt->pending, 0);
+
+       hrtimer_start(&pt->timer, ktime_add_ns(ktime_get(), interval),
+                     HRTIMER_MODE_ABS);
+}
+
+static void pit_load_count(struct kvm *kvm, int channel, u32 val)
+{
+       struct kvm_kpit_state *ps = &kvm->arch.vpit->pit_state;
+
+       WARN_ON(!mutex_is_locked(&ps->lock));
+
+       pr_debug("pit: load_count val is %d, channel is %d\n", val, channel);
+
+       /*
+        * Though spec said the state of 8254 is undefined after power-up,
+        * seems some tricky OS like Windows XP depends on IRQ0 interrupt
+        * when booting up.
+        * So here setting initialize rate for it, and not a specific number
+        */
+       if (val == 0)
+               val = 0x10000;
+
+       ps->channels[channel].count_load_time = ktime_get();
+       ps->channels[channel].count = val;
+
+       if (channel != 0)
+               return;
+
+       /* Two types of timer
+        * mode 1 is one shot, mode 2 is period, otherwise del timer */
+       switch (ps->channels[0].mode) {
+       case 1:
+               create_pit_timer(&ps->pit_timer, val, 0);
+               break;
+       case 2:
+               create_pit_timer(&ps->pit_timer, val, 1);
+               break;
+       default:
+               destroy_pit_timer(&ps->pit_timer);
+       }
+}
+
+static void pit_ioport_write(struct kvm_io_device *this,
+                            gpa_t addr, int len, const void *data)
+{
+       struct kvm_pit *pit = (struct kvm_pit *)this->private;
+       struct kvm_kpit_state *pit_state = &pit->pit_state;
+       struct kvm *kvm = pit->kvm;
+       int channel, access;
+       struct kvm_kpit_channel_state *s;
+       u32 val = *(u32 *) data;
+
+       val  &= 0xff;
+       addr &= KVM_PIT_CHANNEL_MASK;
+
+       mutex_lock(&pit_state->lock);
+
+       if (val != 0)
+               pr_debug("pit: write addr is 0x%x, len is %d, val is 0x%x\n",
+                         (unsigned int)addr, len, val);
+
+       if (addr == 3) {
+               channel = val >> 6;
+               if (channel == 3) {
+                       /* Read-Back Command. */
+                       for (channel = 0; channel < 3; channel++) {
+                               s = &pit_state->channels[channel];
+                               if (val & (2 << channel)) {
+                                       if (!(val & 0x20))
+                                               pit_latch_count(kvm, channel);
+                                       if (!(val & 0x10))
+                                               pit_latch_status(kvm, channel);
+                               }
+                       }
+               } else {
+                       /* Select Counter <channel>. */
+                       s = &pit_state->channels[channel];
+                       access = (val >> 4) & KVM_PIT_CHANNEL_MASK;
+                       if (access == 0) {
+                               pit_latch_count(kvm, channel);
+                       } else {
+                               s->rw_mode = access;
+                               s->read_state = access;
+                               s->write_state = access;
+                               s->mode = (val >> 1) & 7;
+                               if (s->mode > 5)
+                                       s->mode -= 4;
+                               s->bcd = val & 1;
+                       }
+               }
+       } else {
+               /* Write Count. */
+               s = &pit_state->channels[addr];
+               switch (s->write_state) {
+               default:
+               case RW_STATE_LSB:
+                       pit_load_count(kvm, addr, val);
+                       break;
+               case RW_STATE_MSB:
+                       pit_load_count(kvm, addr, val << 8);
+                       break;
+               case RW_STATE_WORD0:
+                       s->write_latch = val;
+                       s->write_state = RW_STATE_WORD1;
+                       break;
+               case RW_STATE_WORD1:
+                       pit_load_count(kvm, addr, s->write_latch | (val << 8));
+                       s->write_state = RW_STATE_WORD0;
+                       break;
+               }
+       }
+
+       mutex_unlock(&pit_state->lock);
+}
+
+static void pit_ioport_read(struct kvm_io_device *this,
+                           gpa_t addr, int len, void *data)
+{
+       struct kvm_pit *pit = (struct kvm_pit *)this->private;
+       struct kvm_kpit_state *pit_state = &pit->pit_state;
+       struct kvm *kvm = pit->kvm;
+       int ret, count;
+       struct kvm_kpit_channel_state *s;
+
+       addr &= KVM_PIT_CHANNEL_MASK;
+       s = &pit_state->channels[addr];
+
+       mutex_lock(&pit_state->lock);
+
+       if (s->status_latched) {
+               s->status_latched = 0;
+               ret = s->status;
+       } else if (s->count_latched) {
+               switch (s->count_latched) {
+               default:
+               case RW_STATE_LSB:
+                       ret = s->latched_count & 0xff;
+                       s->count_latched = 0;
+                       break;
+               case RW_STATE_MSB:
+                       ret = s->latched_count >> 8;
+                       s->count_latched = 0;
+                       break;
+               case RW_STATE_WORD0:
+                       ret = s->latched_count & 0xff;
+                       s->count_latched = RW_STATE_MSB;
+                       break;
+               }
+       } else {
+               switch (s->read_state) {
+               default:
+               case RW_STATE_LSB:
+                       count = pit_get_count(kvm, addr);
+                       ret = count & 0xff;
+                       break;
+               case RW_STATE_MSB:
+                       count = pit_get_count(kvm, addr);
+                       ret = (count >> 8) & 0xff;
+                       break;
+               case RW_STATE_WORD0:
+                       count = pit_get_count(kvm, addr);
+                       ret = count & 0xff;
+                       s->read_state = RW_STATE_WORD1;
+                       break;
+               case RW_STATE_WORD1:
+                       count = pit_get_count(kvm, addr);
+                       ret = (count >> 8) & 0xff;
+                       s->read_state = RW_STATE_WORD0;
+                       break;
+               }
+       }
+
+       if (len > sizeof(ret))
+               len = sizeof(ret);
+       memcpy(data, (char *)&ret, len);
+
+       mutex_unlock(&pit_state->lock);
+}
+
+static int pit_in_range(struct kvm_io_device *this, gpa_t addr)
+{
+       return ((addr >= KVM_PIT_BASE_ADDRESS) &&
+               (addr < KVM_PIT_BASE_ADDRESS + KVM_PIT_MEM_LENGTH));
+}
+
+static void speaker_ioport_write(struct kvm_io_device *this,
+                                gpa_t addr, int len, const void *data)
+{
+       struct kvm_pit *pit = (struct kvm_pit *)this->private;
+       struct kvm_kpit_state *pit_state = &pit->pit_state;
+       struct kvm *kvm = pit->kvm;
+       u32 val = *(u32 *) data;
+
+       mutex_lock(&pit_state->lock);
+       pit_state->speaker_data_on = (val >> 1) & 1;
+       pit_set_gate(kvm, 2, val & 1);
+       mutex_unlock(&pit_state->lock);
+}
+
+static void speaker_ioport_read(struct kvm_io_device *this,
+                               gpa_t addr, int len, void *data)
+{
+       struct kvm_pit *pit = (struct kvm_pit *)this->private;
+       struct kvm_kpit_state *pit_state = &pit->pit_state;
+       struct kvm *kvm = pit->kvm;
+       unsigned int refresh_clock;
+       int ret;
+
+       /* Refresh clock toggles at about 15us. We approximate as 2^14ns. */
+       refresh_clock = ((unsigned int)ktime_to_ns(ktime_get()) >> 14) & 1;
+
+       mutex_lock(&pit_state->lock);
+       ret = ((pit_state->speaker_data_on << 1) | pit_get_gate(kvm, 2) |
+               (pit_get_out(kvm, 2) << 5) | (refresh_clock << 4));
+       if (len > sizeof(ret))
+               len = sizeof(ret);
+       memcpy(data, (char *)&ret, len);
+       mutex_unlock(&pit_state->lock);
+}
+
+static int speaker_in_range(struct kvm_io_device *this, gpa_t addr)
+{
+       return (addr == KVM_SPEAKER_BASE_ADDRESS);
+}
+
+struct kvm_pit *kvm_create_pit(struct kvm *kvm)
+{
+       int i;
+       struct kvm_pit *pit;
+       struct kvm_kpit_state *pit_state;
+       struct kvm_kpit_channel_state *c;
+
+       pit = kzalloc(sizeof(struct kvm_pit), GFP_KERNEL);
+       if (!pit)
+               return NULL;
+
+       mutex_init(&pit->pit_state.lock);
+       mutex_lock(&pit->pit_state.lock);
+
+       /* Initialize PIO device */
+       pit->dev.read = pit_ioport_read;
+       pit->dev.write = pit_ioport_write;
+       pit->dev.in_range = pit_in_range;
+       pit->dev.private = pit;
+       kvm_io_bus_register_dev(&kvm->pio_bus, &pit->dev);
+
+       pit->speaker_dev.read = speaker_ioport_read;
+       pit->speaker_dev.write = speaker_ioport_write;
+       pit->speaker_dev.in_range = speaker_in_range;
+       pit->speaker_dev.private = pit;
+       kvm_io_bus_register_dev(&kvm->pio_bus, &pit->speaker_dev);
+
+       kvm->arch.vpit = pit;
+       pit->kvm = kvm;
+
+       pit_state = &pit->pit_state;
+       pit_state->pit = pit;
+       hrtimer_init(&pit_state->pit_timer.timer,
+                    CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
+       atomic_set(&pit_state->pit_timer.pending, 0);
+       for (i = 0; i < 3; i++) {
+               c = &pit_state->channels[i];
+               c->mode = 0xff;
+               c->gate = (i != 2);
+               pit_load_count(kvm, i, 0);
+       }
+
+       mutex_unlock(&pit->pit_state.lock);
+
+       pit->pit_state.inject_pending = 1;
+
+       return pit;
+}
+
+void kvm_free_pit(struct kvm *kvm)
+{
+       struct hrtimer *timer;
+
+       if (kvm->arch.vpit) {
+               mutex_lock(&kvm->arch.vpit->pit_state.lock);
+               timer = &kvm->arch.vpit->pit_state.pit_timer.timer;
+               hrtimer_cancel(timer);
+               mutex_unlock(&kvm->arch.vpit->pit_state.lock);
+               kfree(kvm->arch.vpit);
+       }
+}
+
+void __inject_pit_timer_intr(struct kvm *kvm)
+{
+       mutex_lock(&kvm->lock);
+       kvm_ioapic_set_irq(kvm->arch.vioapic, 0, 1);
+       kvm_ioapic_set_irq(kvm->arch.vioapic, 0, 0);
+       kvm_pic_set_irq(pic_irqchip(kvm), 0, 1);
+       kvm_pic_set_irq(pic_irqchip(kvm), 0, 0);
+       mutex_unlock(&kvm->lock);
+}
+
+void kvm_inject_pit_timer_irqs(struct kvm_vcpu *vcpu)
+{
+       struct kvm_pit *pit = vcpu->kvm->arch.vpit;
+       struct kvm *kvm = vcpu->kvm;
+       struct kvm_kpit_state *ps;
+
+       if (vcpu && pit) {
+               ps = &pit->pit_state;
+
+               /* Try to inject pending interrupts when:
+                * 1. Pending exists
+                * 2. Last interrupt was accepted or waited for too long time*/
+               if (atomic_read(&ps->pit_timer.pending) &&
+                   (ps->inject_pending ||
+                   (jiffies - ps->last_injected_time
+                               >= KVM_MAX_PIT_INTR_INTERVAL))) {
+                       ps->inject_pending = 0;
+                       __inject_pit_timer_intr(kvm);
+                       ps->last_injected_time = jiffies;
+               }
+       }
+}
+
+void kvm_pit_timer_intr_post(struct kvm_vcpu *vcpu, int vec)
+{
+       struct kvm_arch *arch = &vcpu->kvm->arch;
+       struct kvm_kpit_state *ps;
+
+       if (vcpu && arch->vpit) {
+               ps = &arch->vpit->pit_state;
+               if (atomic_read(&ps->pit_timer.pending) &&
+               (((arch->vpic->pics[0].imr & 1) == 0 &&
+                 arch->vpic->pics[0].irq_base == vec) ||
+                 (arch->vioapic->redirtbl[0].fields.vector == vec &&
+                 arch->vioapic->redirtbl[0].fields.mask != 1))) {
+                       ps->inject_pending = 1;
+                       atomic_dec(&ps->pit_timer.pending);
+                       ps->channels[0].count_load_time = ktime_get();
+               }
+       }
+}
diff --git a/arch/x86/kvm/i8254.h b/arch/x86/kvm/i8254.h
new file mode 100644 (file)
index 0000000..d77d6b7
--- /dev/null
@@ -0,0 +1,61 @@
+#ifndef __I8254_H
+#define __I8254_H
+
+#include "iodev.h"
+
+struct kvm_kpit_timer {
+       struct hrtimer timer;
+       int irq;
+       s64 period; /* unit: ns */
+       s64 scheduled;
+       ktime_t last_update;
+       atomic_t pending;
+};
+
+struct kvm_kpit_channel_state {
+       u32 count; /* can be 65536 */
+       u16 latched_count;
+       u8 count_latched;
+       u8 status_latched;
+       u8 status;
+       u8 read_state;
+       u8 write_state;
+       u8 write_latch;
+       u8 rw_mode;
+       u8 mode;
+       u8 bcd; /* not supported */
+       u8 gate; /* timer start */
+       ktime_t count_load_time;
+};
+
+struct kvm_kpit_state {
+       struct kvm_kpit_channel_state channels[3];
+       struct kvm_kpit_timer pit_timer;
+       u32    speaker_data_on;
+       struct mutex lock;
+       struct kvm_pit *pit;
+       bool inject_pending; /* if inject pending interrupts */
+       unsigned long last_injected_time;
+};
+
+struct kvm_pit {
+       unsigned long base_addresss;
+       struct kvm_io_device dev;
+       struct kvm_io_device speaker_dev;
+       struct kvm *kvm;
+       struct kvm_kpit_state pit_state;
+};
+
+#define KVM_PIT_BASE_ADDRESS       0x40
+#define KVM_SPEAKER_BASE_ADDRESS    0x61
+#define KVM_PIT_MEM_LENGTH         4
+#define KVM_PIT_FREQ               1193181
+#define KVM_MAX_PIT_INTR_INTERVAL   HZ / 100
+#define KVM_PIT_CHANNEL_MASK       0x3
+
+void kvm_inject_pit_timer_irqs(struct kvm_vcpu *vcpu);
+void kvm_pit_timer_intr_post(struct kvm_vcpu *vcpu, int vec);
+struct kvm_pit *kvm_create_pit(struct kvm *kvm);
+void kvm_free_pit(struct kvm *kvm);
+
+#endif
index e571475..dbfe21c 100644 (file)
@@ -23,6 +23,7 @@
 #include <linux/kvm_host.h>
 
 #include "irq.h"
+#include "i8254.h"
 
 /*
  * check if there is pending interrupt without
@@ -66,6 +67,7 @@ EXPORT_SYMBOL_GPL(kvm_cpu_get_interrupt);
 void kvm_inject_pending_timer_irqs(struct kvm_vcpu *vcpu)
 {
        kvm_inject_apic_timer_irqs(vcpu);
+       kvm_inject_pit_timer_irqs(vcpu);
        /* TODO: PIT, RTC etc. */
 }
 EXPORT_SYMBOL_GPL(kvm_inject_pending_timer_irqs);
@@ -73,6 +75,7 @@ EXPORT_SYMBOL_GPL(kvm_inject_pending_timer_irqs);
 void kvm_timer_intr_post(struct kvm_vcpu *vcpu, int vec)
 {
        kvm_apic_timer_intr_post(vcpu, vec);
+       kvm_pit_timer_intr_post(vcpu, vec);
        /* TODO: PIT, RTC etc. */
 }
 EXPORT_SYMBOL_GPL(kvm_timer_intr_post);
index bf78d65..c33a457 100644 (file)
@@ -17,6 +17,7 @@
 #include <linux/kvm_host.h>
 #include "irq.h"
 #include "mmu.h"
+#include "i8254.h"
 
 #include <linux/clocksource.h>
 #include <linux/kvm.h>
@@ -818,6 +819,7 @@ int kvm_dev_ioctl_check_extension(long ext)
        case KVM_CAP_SET_TSS_ADDR:
        case KVM_CAP_EXT_CPUID:
        case KVM_CAP_CLOCKSOURCE:
+       case KVM_CAP_PIT:
                r = 1;
                break;
        case KVM_CAP_VAPIC:
@@ -1594,6 +1596,12 @@ long kvm_arch_vm_ioctl(struct file *filp,
                } else
                        goto out;
                break;
+       case KVM_CREATE_PIT:
+               r = -ENOMEM;
+               kvm->arch.vpit = kvm_create_pit(kvm);
+               if (kvm->arch.vpit)
+                       r = 0;
+               break;
        case KVM_IRQ_LINE: {
                struct kvm_irq_level irq_event;
 
@@ -3372,6 +3380,7 @@ static void kvm_free_vcpus(struct kvm *kvm)
 
 void kvm_arch_destroy_vm(struct kvm *kvm)
 {
+       kvm_free_pit(kvm);
        kfree(kvm->arch.vpic);
        kfree(kvm->arch.vioapic);
        kvm_free_vcpus(kvm);
index 49ced21..26a313a 100644 (file)
@@ -298,6 +298,7 @@ struct kvm_arch{
        struct list_head active_mmu_pages;
        struct kvm_pic *vpic;
        struct kvm_ioapic *vioapic;
+       struct kvm_pit *vpit;
 
        int round_robin_prev_vcpu;
        unsigned int tss_addr;
index e92e703..cefa9a2 100644 (file)
@@ -236,6 +236,7 @@ struct kvm_vapic_addr {
 #define KVM_CAP_CLOCKSOURCE 8
 #define KVM_CAP_NR_VCPUS 9       /* returns max vcpus per vm */
 #define KVM_CAP_NR_MEMSLOTS 10   /* returns max memory slots per vm */
+#define KVM_CAP_PIT 11
 
 /*
  * ioctls for VM fds
@@ -258,6 +259,7 @@ struct kvm_vapic_addr {
 #define KVM_IRQ_LINE             _IOW(KVMIO, 0x61, struct kvm_irq_level)
 #define KVM_GET_IRQCHIP                  _IOWR(KVMIO, 0x62, struct kvm_irqchip)
 #define KVM_SET_IRQCHIP                  _IOR(KVMIO,  0x63, struct kvm_irqchip)
+#define KVM_CREATE_PIT           _IO(KVMIO,  0x64)
 
 /*
  * ioctls for vcpu fds