Taking the distributor lock in the VGIC maintainance handler
has the unfortunate consequence of forcing all the other users
of that lock to disable interrupts while this lock is held.
This lock is only held to update the irq_active bit for an EOIed
level interrupt. If we remove the locking, there is a race with
the queuing of an interrupt in __kvm_sync_to_cpu(), where we check
if the interrupt is already active. Two possibilities:
- The queuing is occuring on the same vcpu: cannot happen, as we're
already in the context of this vcpu, and executing the handler
- The interrupt has been migrated to another vcpu, and we ignore
this interrupt for this run. Big deal. It is still pending though,
and will get considered when this vcpu exits.
We can then safely remove that lock from vgic_maintainance_handler()
and convert the other users not to disable interrupts.
Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx>
---
arch/arm/kvm/vgic.c | 25 ++++++++++---------------
1 file changed, 10 insertions(+), 15 deletions(-)
diff --git a/arch/arm/kvm/vgic.c b/arch/arm/kvm/vgic.c
index fd8865d..b77f8bf 100644
--- a/arch/arm/kvm/vgic.c
+++ b/arch/arm/kvm/vgic.c
@@ -549,9 +549,9 @@ bool vgic_handle_mmio(struct kvm_vcpu *vcpu, struct kvm_run *run, struct kvm_exi
spin_lock(&vcpu->kvm->arch.vgic.lock);
updated_state = range->handle_mmio(vcpu, mmio,mmio->phys_addr - range->base - base);
+ spin_unlock(&vcpu->kvm->arch.vgic.lock);
kvm_prepare_mmio(run, mmio);
kvm_handle_mmio_return(vcpu, run);
- spin_unlock(&vcpu->kvm->arch.vgic.lock);
if (updated_state)
vgic_kick_vcpus(vcpu->kvm);
@@ -822,27 +822,25 @@ static void __kvm_vgic_sync_from_cpu(struct kvm_vcpu *vcpu)
void kvm_vgic_sync_to_cpu(struct kvm_vcpu *vcpu)
{
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
- unsigned long flags;
if (!irqchip_in_kernel(vcpu->kvm))
return;
- spin_lock_irqsave(&dist->lock, flags);
+ spin_lock(&dist->lock);
__kvm_vgic_sync_to_cpu(vcpu);
- spin_unlock_irqrestore(&dist->lock, flags);
+ spin_unlock(&dist->lock);
}
void kvm_vgic_sync_from_cpu(struct kvm_vcpu *vcpu)
{
struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
- unsigned long flags;
if (!irqchip_in_kernel(vcpu->kvm))
return;
- spin_lock_irqsave(&dist->lock, flags);
+ spin_lock(&dist->lock);
__kvm_vgic_sync_from_cpu(vcpu);
- spin_unlock_irqrestore(&dist->lock, flags);
+ spin_unlock(&dist->lock);
}
int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
@@ -877,9 +875,8 @@ static bool vgic_update_irq_state(struct kvm *kvm, int cpuid,
struct kvm_vcpu *vcpu;
int is_edge, state;
int pend, enabled;
- unsigned long flags;
- spin_lock_irqsave(&dist->lock, flags);
+ spin_lock(&dist->lock);
is_edge = vgic_irq_is_edge(dist, irq_num);
state = vgic_bitmap_get_irq_val(&dist->irq_state, cpuid, irq_num);
@@ -889,9 +886,8 @@ static bool vgic_update_irq_state(struct kvm *kvm, int cpuid,
* - level triggered and we change level
* - edge triggered and we have a rising edge
*/
- if (!((!is_edge && (state ^ level)) ||
- (is_edge && !state && level))) {
- spin_unlock_irqrestore(&dist->lock, flags);
+ if (!((!is_edge && (state ^ level)) || (is_edge && !state && level))) {
+ spin_unlock(&dist->lock);
return false;
}
@@ -915,7 +911,7 @@ static bool vgic_update_irq_state(struct kvm *kvm, int cpuid,
} else
clear_bit(irq_num, vcpu->arch.vgic_cpu.pending);
- spin_unlock_irqrestore(&dist->lock, flags);
+ spin_unlock(&dist->lock);
return true;
}
@@ -950,7 +946,6 @@ static irqreturn_t vgic_maintainance_handler(int irq, void *data)
*/
int lr, irq;
- spin_lock(&dist->lock);
for_each_set_bit(lr, (unsigned long *)vgic_cpu->vgic_eisr,
vgic_cpu->nr_lr) {
irq = vgic_cpu->vgic_lr[lr] & VGIC_LR_VIRTUALID;
@@ -958,11 +953,11 @@ static irqreturn_t vgic_maintainance_handler(int irq, void *data)
vgic_bitmap_set_irq_val(&dist->irq_active,
vcpu->vcpu_id, irq, 0);
atomic_dec(&vgic_cpu->irq_active_count);
+ smp_mb();
vgic_cpu->vgic_lr[lr] &= ~VGIC_LR_EOI;
writel_relaxed(vgic_cpu->vgic_lr[lr],
dist->vctrl_base + GICH_LR0 + (lr << 2));
}
- spin_unlock(&dist->lock);
}
if (vgic_cpu->vgic_misr & VGIC_MISR_U) {
--
1.7.12
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
