On Thu, Jan 10, 2013 at 07:36:28PM -0200, Marcelo Tosatti wrote:
> Hi,
>
> Getting into good shape.
>
> On Thu, Jan 10, 2013 at 03:26:08PM +0800, Yang Zhang wrote:
> > From: Yang Zhang <[email protected]>
> >
> > Virtual interrupt delivery avoids KVM to inject vAPIC interrupts
> > manually, which is fully taken care of by the hardware. This needs
> > some special awareness into existing interrupr injection path:
> >
> > - for pending interrupt, instead of direct injection, we may need
> > update architecture specific indicators before resuming to guest.
> >
> > - A pending interrupt, which is masked by ISR, should be also
> > considered in above update action, since hardware will decide
> > when to inject it at right time. Current has_interrupt and
> > get_interrupt only returns a valid vector from injection p.o.v.
> >
> > Signed-off-by: Kevin Tian <[email protected]>
> > Signed-off-by: Yang Zhang <[email protected]>
> > ---
> > arch/x86/include/asm/kvm_host.h | 5 +
> > arch/x86/include/asm/vmx.h | 11 +++
> > arch/x86/kvm/irq.c | 56 +++++++++++-
> > arch/x86/kvm/lapic.c | 72 +++++++++------
> > arch/x86/kvm/lapic.h | 23 +++++
> > arch/x86/kvm/svm.c | 18 ++++
> > arch/x86/kvm/vmx.c | 191
> > +++++++++++++++++++++++++++++++++++++--
> > arch/x86/kvm/x86.c | 14 +++-
> > include/linux/kvm_host.h | 3 +
> > virt/kvm/ioapic.c | 18 ++++
> > virt/kvm/ioapic.h | 4 +
> > virt/kvm/irq_comm.c | 22 +++++
> > virt/kvm/kvm_main.c | 5 +
> > 13 files changed, 399 insertions(+), 43 deletions(-)
> >
>
> > static void recalculate_apic_map(struct kvm *kvm)
> > {
> > struct kvm_apic_map *new, *old = NULL;
> > @@ -236,12 +219,14 @@ static inline void kvm_apic_set_id(struct kvm_lapic
> > *apic, u8 id)
> > {
> > apic_set_reg(apic, APIC_ID, id << 24);
> > recalculate_apic_map(apic->vcpu->kvm);
> > + ioapic_update_eoi_exitmap(apic->vcpu->kvm);
> > }
>
> Move ioapic_update_eoi_exitmap into recalculate_apic_map.
>
> > static inline void kvm_apic_set_ldr(struct kvm_lapic *apic, u32 id)
> > {
> > apic_set_reg(apic, APIC_LDR, id);
> > recalculate_apic_map(apic->vcpu->kvm);
> > + ioapic_update_eoi_exitmap(apic->vcpu->kvm);
> > }
> >
>
> > diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> > index b203ce7..990409a 100644
> > --- a/arch/x86/kvm/vmx.c
> > +++ b/arch/x86/kvm/vmx.c
> > @@ -434,6 +434,7 @@ struct vcpu_vmx {
> > bool rdtscp_enabled;
> >
> > bool virtual_x2apic_enabled;
> > + unsigned long eoi_exit_bitmap[4];
>
> Use DECLARE_BITMAP (unsigned long is 4 bytes on 32-bit host).
>
> > /* Support for a guest hypervisor (nested VMX) */
> > struct nested_vmx nested;
> > @@ -783,7 +784,8 @@ static inline bool cpu_has_vmx_apic_register_virt(void)
> >
> > static inline bool cpu_has_vmx_virtual_intr_delivery(void)
> > {
> > - return false;
> > + return vmcs_config.cpu_based_2nd_exec_ctrl &
> > + SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY;
> > }
> >
> > static inline bool cpu_has_vmx_flexpriority(void)
> > @@ -2565,7 +2567,8 @@ static __init int setup_vmcs_config(struct
> > vmcs_config *vmcs_conf)
> > SECONDARY_EXEC_PAUSE_LOOP_EXITING |
> > SECONDARY_EXEC_RDTSCP |
> > SECONDARY_EXEC_ENABLE_INVPCID |
> > - SECONDARY_EXEC_APIC_REGISTER_VIRT;
> > + SECONDARY_EXEC_APIC_REGISTER_VIRT |
> > + SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY;
> > if (adjust_vmx_controls(min2, opt2,
> > MSR_IA32_VMX_PROCBASED_CTLS2,
> > &_cpu_based_2nd_exec_control) < 0)
> > @@ -2579,7 +2582,8 @@ static __init int setup_vmcs_config(struct
> > vmcs_config *vmcs_conf)
> >
> > if (!(_cpu_based_exec_control & CPU_BASED_TPR_SHADOW))
> > _cpu_based_2nd_exec_control &= ~(
> > - SECONDARY_EXEC_APIC_REGISTER_VIRT);
> > + SECONDARY_EXEC_APIC_REGISTER_VIRT |
> > + SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY);
>
> Nevermind the earlier comment about ().
>
> > +static void set_eoi_exitmap_one(struct kvm_vcpu *vcpu,
> > + u32 vector)
> > +{
> > + struct vcpu_vmx *vmx = to_vmx(vcpu);
> > +
> > + if (WARN_ONCE((vector > 255),
> > + "KVM VMX: vector (%d) out of range\n", vector))
> > + return;
> > + __set_bit(vector, vmx->eoi_exit_bitmap);
> > +}
> > +
> > +void vmx_check_ioapic_entry(struct kvm_vcpu *vcpu, struct kvm_lapic_irq
> > *irq)
> > +{
> > + struct kvm_lapic **dst;
> > + struct kvm_apic_map *map;
> > + unsigned long bitmap = 1;
> > + int i;
> > +
> > + rcu_read_lock();
> > + map = rcu_dereference(vcpu->kvm->arch.apic_map);
> > +
> > + if (unlikely(!map)) {
> > + set_eoi_exitmap_one(vcpu, irq->vector);
> > + goto out;
> > + }
> > +
> > + if (irq->dest_mode == 0) { /* physical mode */
> > + if (irq->delivery_mode == APIC_DM_LOWEST ||
> > + irq->dest_id == 0xff) {
> > + set_eoi_exitmap_one(vcpu, irq->vector);
> > + goto out;
> > + }
> > + dst = &map->phys_map[irq->dest_id & 0xff];
> > + } else {
> > + u32 mda = irq->dest_id << (32 - map->ldr_bits);
> > +
> > + dst = map->logical_map[apic_cluster_id(map, mda)];
> > +
> > + bitmap = apic_logical_id(map, mda);
> > + }
> > +
> > + for_each_set_bit(i, &bitmap, 16) {
> > + if (!dst[i])
> > + continue;
> > + if (dst[i]->vcpu == vcpu) {
> > + set_eoi_exitmap_one(vcpu, irq->vector);
> > + break;
> > + }
> > + }
> > +
> > +out:
> > + rcu_read_unlock();
> > +}
> > +
> > +static void vmx_load_eoi_exitmap(struct kvm_vcpu *vcpu)
> > +{
> > + struct vcpu_vmx *vmx = to_vmx(vcpu);
> > +
> > + vmcs_write64(EOI_EXIT_BITMAP0, vmx->eoi_exit_bitmap[0]);
> > + vmcs_write64(EOI_EXIT_BITMAP1, vmx->eoi_exit_bitmap[1]);
> > + vmcs_write64(EOI_EXIT_BITMAP2, vmx->eoi_exit_bitmap[2]);
> > + vmcs_write64(EOI_EXIT_BITMAP3, vmx->eoi_exit_bitmap[3]);
> > +}
> > +
> > +static void vmx_update_eoi_exitmap(struct kvm_vcpu *vcpu)
> > +{
> > + struct kvm_ioapic *ioapic = vcpu->kvm->arch.vioapic;
> > + union kvm_ioapic_redirect_entry *e;
> > + struct kvm_lapic_irq irqe;
> > + int index;
> > + struct vcpu_vmx *vmx = to_vmx(vcpu);
> > +
> > + /* clear eoi exit bitmap */
> > + memset(vmx->eoi_exit_bitmap, 0, 32);
> > +
> > + /* traverse ioapic entry to set eoi exit bitmap*/
> > + for (index = 0; index < IOAPIC_NUM_PINS; index++) {
> > + e = &ioapic->redirtbl[index];
> > + if (!e->fields.mask &&
> > + (e->fields.trig_mode == IOAPIC_LEVEL_TRIG ||
> > + kvm_irq_has_notifier(ioapic->kvm, KVM_IRQCHIP_IOAPIC,
> > + index))) {
> > + irqe.dest_id = e->fields.dest_id;
> > + irqe.vector = e->fields.vector;
> > + irqe.dest_mode = e->fields.dest_mode;
> > + irqe.delivery_mode = e->fields.delivery_mode << 8;
> > + vmx_check_ioapic_entry(vcpu, &irqe);
> > +
> > + }
> > + }
> > +
> > + vmx_load_eoi_exitmap(vcpu);
> > +}
> > +
> > static void vmx_complete_atomic_exit(struct vcpu_vmx *vmx)
> > {
> > u32 exit_intr_info;
> > @@ -7553,6 +7726,10 @@ static struct kvm_x86_ops vmx_x86_ops = {
> > .update_cr8_intercept = update_cr8_intercept,
> > .enable_virtual_x2apic_mode = vmx_enable_virtual_x2apic_mode,
> > .disable_virtual_x2apic_mode = vmx_disable_virtual_x2apic_mode,
> > + .has_virtual_interrupt_delivery = vmx_has_virtual_interrupt_delivery,
> > + .update_apic_irq = vmx_update_apic_irq,
> > + .update_eoi_exitmap = vmx_update_eoi_exitmap,
> > + .set_svi = vmx_set_svi,
> >
> > .set_tss_addr = vmx_set_tss_addr,
> > .get_tdp_level = get_ept_level,
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index 1c9c834..e6d8227 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -5527,7 +5527,7 @@ static void inject_pending_event(struct kvm_vcpu
> > *vcpu)
> > vcpu->arch.nmi_injected = true;
> > kvm_x86_ops->set_nmi(vcpu);
> > }
> > - } else if (kvm_cpu_has_interrupt(vcpu)) {
> > + } else if (kvm_cpu_has_injectable_intr(vcpu)) {
> > if (kvm_x86_ops->interrupt_allowed(vcpu)) {
> > kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu),
> > false);
> > @@ -5648,6 +5648,11 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> > kvm_handle_pmu_event(vcpu);
> > if (kvm_check_request(KVM_REQ_PMI, vcpu))
> > kvm_deliver_pmi(vcpu);
> > + if (kvm_check_request(KVM_REQ_EOIBITMAP, vcpu)) {
> > + mutex_lock(&vcpu->kvm->arch.vioapic->eoimap_lock);
> > + kvm_x86_ops->update_eoi_exitmap(vcpu);
> > + mutex_unlock(&vcpu->kvm->arch.vioapic->eoimap_lock);
> > + }
>
> Take ioapic lock and irq_lock mutex.
>
Why irq_lock? Ack notifiers are RCU protected and they are only read
here.
> > +void ioapic_update_eoi_exitmap(struct kvm *kvm)
> > +{
> > +#ifdef CONFIG_X86
> > + struct kvm_vcpu *vcpu = kvm->vcpus[0];
> > + struct kvm_ioapic *ioapic = kvm->arch.vioapic;
> > +
> > + /* If vid is enabled in one of vcpus, then other
> > + * vcpus also enabled it. */
> > + if (!kvm_apic_vid_enabled(vcpu) || !ioapic)
> > + return;
>
> Is it even possible to call ioapic_update_eoi_exitmap() if
> kvm->arch.vioapic == NULL?
>
May be if irq chip in userspace. Need to verify all callers.
> > + kvm_make_update_eoibitmap_request(kvm);
> > +#endif
> > +}
> > +
> > static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
> > {
> > unsigned index;
> > @@ -156,6 +170,7 @@ static void ioapic_write_indirect(struct kvm_ioapic
> > *ioapic, u32 val)
> > if (e->fields.trig_mode == IOAPIC_LEVEL_TRIG
> > && ioapic->irr & (1 << index))
> > ioapic_service(ioapic, index);
> > + ioapic_update_eoi_exitmap(ioapic->kvm);
> > break;
> > }
> > }
> > @@ -415,6 +430,9 @@ int kvm_ioapic_init(struct kvm *kvm)
> > ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, ioapic->base_address,
> > IOAPIC_MEM_LENGTH, &ioapic->dev);
> > mutex_unlock(&kvm->slots_lock);
> > +#ifdef CONFIG_X86
> > + mutex_init(&ioapic->eoimap_lock);
> > +#endif
> > if (ret < 0) {
> > kvm->arch.vioapic = NULL;
> > kfree(ioapic);
> > diff --git a/virt/kvm/ioapic.h b/virt/kvm/ioapic.h
> > index a30abfe..34544ce 100644
> > --- a/virt/kvm/ioapic.h
> > +++ b/virt/kvm/ioapic.h
> > @@ -47,6 +47,9 @@ struct kvm_ioapic {
> > void (*ack_notifier)(void *opaque, int irq);
> > spinlock_t lock;
> > DECLARE_BITMAP(handled_vectors, 256);
> > +#ifdef CONFIG_X86
> > + struct mutex eoimap_lock;
> > +#endif
> > };
> >
> > #ifdef DEBUG
> > @@ -82,5 +85,6 @@ int kvm_irq_delivery_to_apic(struct kvm *kvm, struct
> > kvm_lapic *src,
> > struct kvm_lapic_irq *irq);
> > int kvm_get_ioapic(struct kvm *kvm, struct kvm_ioapic_state *state);
> > int kvm_set_ioapic(struct kvm *kvm, struct kvm_ioapic_state *state);
> > +void ioapic_update_eoi_exitmap(struct kvm *kvm);
> >
> > #endif
> > diff --git a/virt/kvm/irq_comm.c b/virt/kvm/irq_comm.c
> > index 656fa45..64aa1ab 100644
> > --- a/virt/kvm/irq_comm.c
> > +++ b/virt/kvm/irq_comm.c
> > @@ -22,6 +22,7 @@
> >
> > #include <linux/kvm_host.h>
> > #include <linux/slab.h>
> > +#include <linux/export.h>
>
> Whats this for?
>
> > #include <trace/events/kvm.h>
> >
> > #include <asm/msidef.h>
> > @@ -237,6 +238,25 @@ int kvm_set_irq_inatomic(struct kvm *kvm, int
> > irq_source_id, u32 irq, int level)
> > return ret;
> > }
> >
> > +bool kvm_irq_has_notifier(struct kvm *kvm, unsigned irqchip, unsigned pin)
> > +{
> > + struct kvm_irq_ack_notifier *kian;
> > + struct hlist_node *n;
> > + int gsi;
> > +
> > + rcu_read_lock();
> > + gsi = rcu_dereference(kvm->irq_routing)->chip[irqchip][pin];
> > + if (gsi != -1)
> > + hlist_for_each_entry_rcu(kian, n, &kvm->irq_ack_notifier_list,
> > + link)
> > + if (kian->gsi == gsi)
> > + return true;
>
> Forgot rcu_read_unlock();
>
> > + rcu_read_unlock();
> > +
> > + return false;
> > +}
> > +EXPORT_SYMBOL_GPL(kvm_irq_has_notifier);
> > +
> > void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin)
> > {
> > struct kvm_irq_ack_notifier *kian;
> > @@ -261,6 +281,7 @@ void kvm_register_irq_ack_notifier(struct kvm *kvm,
> > mutex_lock(&kvm->irq_lock);
> > hlist_add_head_rcu(&kian->link, &kvm->irq_ack_notifier_list);
> > mutex_unlock(&kvm->irq_lock);
> > + ioapic_update_eoi_exitmap(kvm);
> > }
>
> Move inside irq_lock protection.
>
> >
> > void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
> > @@ -270,6 +291,7 @@ void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
> > hlist_del_init_rcu(&kian->link);
> > mutex_unlock(&kvm->irq_lock);
> > synchronize_rcu();
> > + ioapic_update_eoi_exitmap(kvm);
>
> Move both synchronize_rcu and ioapic_update_eoi_exitmap inside irq_lock
> protection.
Why? (here and one above). If vcpu uses stale data during update it will
find recalculate request during guest entry and will recalculate again.
--
Gleb.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
