KVM: x86: Introduce a workqueue to deliver PIT timer interrupts

We really want to "kvm_set_irq" during the hrtimer callback,
but that is risky because that is during interrupt context.
Instead, offload the work to a workqueue, which is a bit safer
and should provide most of the same functionality.

Signed-off-by: Chris Lalancette <clalance@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

1 files changed, 85 insertions, 56 deletions

diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
index 188d82762c1b..467cc47fb733 100644
--- a/arch/x86/kvm/i8254.c
+++ b/arch/x86/kvm/i8254.c
@@ -34,6 +34,7 @@
 
 #include <linux/kvm_host.h>
 #include <linux/slab.h>
+#include <linux/workqueue.h>
 
 #include "irq.h"
 #include "i8254.h"
@@ -244,11 +245,22 @@ static void kvm_pit_ack_irq(struct kvm_irq_ack_notifier *kian)
 {
         struct kvm_kpit_state *ps = container_of(kian, struct kvm_kpit_state,
                                                  irq_ack_notifier);
-        raw_spin_lock(&ps->inject_lock);
-        if (atomic_dec_return(&ps->pit_timer.pending) < 0)
+        int value;
+
+        spin_lock(&ps->inject_lock);
+        value = atomic_dec_return(&ps->pit_timer.pending);
+        if (value < 0)
+                /* spurious acks can be generated if, for example, the
+                 * PIC is being reset.  Handle it gracefully here
+                 */
                 atomic_inc(&ps->pit_timer.pending);
+        else if (value > 0)
+                /* in this case, we had multiple outstanding pit interrupts
+                 * that we needed to inject.  Reinject
+                 */
+                queue_work(ps->pit->wq, &ps->pit->expired);
         ps->irq_ack = 1;
-        raw_spin_unlock(&ps->inject_lock);
+        spin_unlock(&ps->inject_lock);
 }
 
 void __kvm_migrate_pit_timer(struct kvm_vcpu *vcpu)
@@ -264,10 +276,10 @@ void __kvm_migrate_pit_timer(struct kvm_vcpu *vcpu)
                 hrtimer_start_expires(timer, HRTIMER_MODE_ABS);
 }
 
-static void destroy_pit_timer(struct kvm_timer *pt)
+static void destroy_pit_timer(struct kvm_pit *pit)
 {
-        pr_debug("execute del timer!\n");
-        hrtimer_cancel(&pt->timer);
+        hrtimer_cancel(&pit->pit_state.pit_timer.timer);
+        cancel_work_sync(&pit->expired);
 }
 
 static bool kpit_is_periodic(struct kvm_timer *ktimer)
@@ -281,6 +293,60 @@ static struct kvm_timer_ops kpit_ops = {
         .is_periodic = kpit_is_periodic,
 };
 
+static void pit_do_work(struct work_struct *work)
+{
+        struct kvm_pit *pit = container_of(work, struct kvm_pit, expired);
+        struct kvm *kvm = pit->kvm;
+        struct kvm_vcpu *vcpu;
+        int i;
+        struct kvm_kpit_state *ps = &pit->pit_state;
+        int inject = 0;
+
+        /* Try to inject pending interrupts when
+         * last one has been acked.
+         */
+        spin_lock(&ps->inject_lock);
+        if (ps->irq_ack) {
+                ps->irq_ack = 0;
+                inject = 1;
+        }
+        spin_unlock(&ps->inject_lock);
+        if (inject) {
+                kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 1);
+                kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 0);
+
+                /*
+                 * Provides NMI watchdog support via Virtual Wire mode.
+                 * The route is: PIT -> PIC -> LVT0 in NMI mode.
+                 *
+                 * Note: Our Virtual Wire implementation is simplified, only
+                 * propagating PIT interrupts to all VCPUs when they have set
+                 * LVT0 to NMI delivery. Other PIC interrupts are just sent to
+                 * VCPU0, and only if its LVT0 is in EXTINT mode.
+                 */
+                if (kvm->arch.vapics_in_nmi_mode > 0)
+                        kvm_for_each_vcpu(i, vcpu, kvm)
+                                kvm_apic_nmi_wd_deliver(vcpu);
+        }
+}
+
+static enum hrtimer_restart pit_timer_fn(struct hrtimer *data)
+{
+        struct kvm_timer *ktimer = container_of(data, struct kvm_timer, timer);
+        struct kvm_pit *pt = ktimer->kvm->arch.vpit;
+
+        if (ktimer->reinject || !atomic_read(&ktimer->pending)) {
+                atomic_inc(&ktimer->pending);
+                queue_work(pt->wq, &pt->expired);
+        }
+
+        if (ktimer->t_ops->is_periodic(ktimer)) {
+                hrtimer_add_expires_ns(&ktimer->timer, ktimer->period);
+                return HRTIMER_RESTART;
+        } else
+                return HRTIMER_NORESTART;
+}
+
 static void create_pit_timer(struct kvm_kpit_state *ps, u32 val, int is_period)
 {
         struct kvm_timer *pt = &ps->pit_timer;
@@ -292,13 +358,13 @@ static void create_pit_timer(struct kvm_kpit_state *ps, u32 val, int is_period)
 
         /* TODO The new value only affected after the retriggered */
         hrtimer_cancel(&pt->timer);
+        cancel_work_sync(&ps->pit->expired);
         pt->period = interval;
         ps->is_periodic = is_period;
 
-        pt->timer.function = kvm_timer_fn;
+        pt->timer.function = pit_timer_fn;
         pt->t_ops = &kpit_ops;
         pt->kvm = ps->pit->kvm;
-        pt->vcpu = pt->kvm->bsp_vcpu;
 
         atomic_set(&pt->pending, 0);
         ps->irq_ack = 1;
@@ -347,7 +413,7 @@ static void pit_load_count(struct kvm *kvm, int channel, u32 val)
                 }
                 break;
         default:
-                destroy_pit_timer(&ps->pit_timer);
+                destroy_pit_timer(kvm->arch.vpit);
         }
 }
 
@@ -626,7 +692,14 @@ struct kvm_pit *kvm_create_pit(struct kvm *kvm, u32 flags)
 
         mutex_init(&pit->pit_state.lock);
         mutex_lock(&pit->pit_state.lock);
-        raw_spin_lock_init(&pit->pit_state.inject_lock);
+        spin_lock_init(&pit->pit_state.inject_lock);
+
+        pit->wq = create_singlethread_workqueue("kvm-pit-wq");
+        if (!pit->wq) {
+                kfree(pit);
+                return NULL;
+        }
+        INIT_WORK(&pit->expired, pit_do_work);
 
         kvm->arch.vpit = pit;
         pit->kvm = kvm;
@@ -685,54 +758,10 @@ void kvm_free_pit(struct kvm *kvm)
                 mutex_lock(&kvm->arch.vpit->pit_state.lock);
                 timer = &kvm->arch.vpit->pit_state.pit_timer.timer;
                 hrtimer_cancel(timer);
+                cancel_work_sync(&kvm->arch.vpit->expired);
                 kvm_free_irq_source_id(kvm, kvm->arch.vpit->irq_source_id);
                 mutex_unlock(&kvm->arch.vpit->pit_state.lock);
+                destroy_workqueue(kvm->arch.vpit->wq);
                 kfree(kvm->arch.vpit);
         }
 }
-
-static void __inject_pit_timer_intr(struct kvm *kvm)
-{
-        struct kvm_vcpu *vcpu;
-        int i;
-
-        kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 1);
-        kvm_set_irq(kvm, kvm->arch.vpit->irq_source_id, 0, 0);
-
-        /*
-         * Provides NMI watchdog support via Virtual Wire mode.
-         * The route is: PIT -> PIC -> LVT0 in NMI mode.
-         *
-         * Note: Our Virtual Wire implementation is simplified, only
-         * propagating PIT interrupts to all VCPUs when they have set
-         * LVT0 to NMI delivery. Other PIC interrupts are just sent to
-         * VCPU0, and only if its LVT0 is in EXTINT mode.
-         */
-        if (kvm->arch.vapics_in_nmi_mode > 0)
-                kvm_for_each_vcpu(i, vcpu, kvm)
-                        kvm_apic_nmi_wd_deliver(vcpu);
-}
-
-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 (pit) {
-                int inject = 0;
-                ps = &pit->pit_state;
-
-                /* Try to inject pending interrupts when
-                 * last one has been acked.
-                 */
-                raw_spin_lock(&ps->inject_lock);
-                if (atomic_read(&ps->pit_timer.pending) && ps->irq_ack) {
-                        ps->irq_ack = 0;
-                        inject = 1;
-                }
-                raw_spin_unlock(&ps->inject_lock);
-                if (inject)
-                        __inject_pit_timer_intr(kvm);
-        }
-}