[PATCH v4 22/22] KVM: arm64: ITS: Pending table save/restore

[Marc Zyngier]

On 27/03/17 10:31, Eric Auger wrote:
> Save and restore the pending tables.
> 
> Pending table restore obviously requires the pendbaser to be
> already set.
> 
> Signed-off-by: Eric Auger <eric.auger at redhat.com>
> 
> ---
> 
> v3 -> v4:
> - remove the wrong comment about locking
> - pass kvm struct instead of its handle
> - add comment about restore method
> - remove GITR_PENDABASER.PTZ check
> - continue if target_vcpu == NULL
> - new locking strategy
> 
> v1 -> v2:
> - do not care about the 1st KB which should be zeroed according to
>   the spec.
> ---
>  virt/kvm/arm/vgic/vgic-its.c | 66 ++++++++++++++++++++++++++++++++++++++++++--
>  1 file changed, 64 insertions(+), 2 deletions(-)
> 
> diff --git a/virt/kvm/arm/vgic/vgic-its.c b/virt/kvm/arm/vgic/vgic-its.c
> index a516bbb..e10aa81 100644
> --- a/virt/kvm/arm/vgic/vgic-its.c
> +++ b/virt/kvm/arm/vgic/vgic-its.c
> @@ -1804,16 +1804,78 @@ static int lookup_table(struct vgic_its *its, gpa_t base, int size, int esz,
>   */
>  static int vgic_its_flush_pending_tables(struct kvm *kvm)
>  {
> -	return -ENXIO;
> +	struct vgic_dist *dist = &kvm->arch.vgic;
> +	struct vgic_irq *irq;
> +	int ret;
> +
> +	list_for_each_entry(irq, &dist->lpi_list_head, lpi_list) {
> +		struct kvm_vcpu *vcpu;
> +		gpa_t pendbase, ptr;
> +		bool stored;
> +		u8 val;
> +
> +		vcpu = irq->target_vcpu;
> +		if (!vcpu)
> +			continue;
> +
> +		pendbase = PENDBASER_ADDRESS(vcpu->arch.vgic_cpu.pendbaser);
> +
> +		ptr = pendbase + (irq->intid / BITS_PER_BYTE);
> +
> +		ret = kvm_read_guest(kvm, (gpa_t)ptr, &val, 1);

ptr is already a gpa_t...

> +		if (ret)
> +			return ret;
> +
> +		stored = val & (irq->intid % BITS_PER_BYTE);
> +		if (stored == irq->pending_latch)
> +			continue;
> +
> +		if (irq->pending_latch)
> +			val |= 1 << (irq->intid % BITS_PER_BYTE);
> +		else
> +			val &= ~(1 << (irq->intid % BITS_PER_BYTE));
> +
> +		ret = kvm_write_guest(kvm, (gpa_t)ptr, &val, 1);
> +		if (ret)
> +			return ret;

Consider the optimization used in its_sync_lpi_pending_table(), where
consecutive LPIs stored in the same byte are set in a single access.

> +	}
> +
> +	return 0;
>  }
>  
>  /**
>   * vgic_its_restore_pending_tables - Restore the pending tables from guest
>   * RAM to internal data structs
> + *
> + * Does not scan the whole pending tables but just loop on all registered
> + * LPIS and scan their associated pending bit. This obviously requires
> + * the ITEs to be restored before.
>   */
>  static int vgic_its_restore_pending_tables(struct kvm *kvm)
>  {
> -	return -ENXIO;
> +	struct vgic_dist *dist = &kvm->arch.vgic;
> +	struct vgic_irq *irq;
> +	int ret;
> +
> +	list_for_each_entry(irq, &dist->lpi_list_head, lpi_list) {
> +		struct kvm_vcpu *vcpu;
> +		gpa_t pendbase, ptr;
> +		u8 val;
> +
> +		vcpu = irq->target_vcpu;
> +		if (!vcpu)
> +			continue;
> +
> +		pendbase = PENDBASER_ADDRESS(vcpu->arch.vgic_cpu.pendbaser);
> +
> +		ptr = pendbase + (irq->intid / BITS_PER_BYTE);
> +
> +		ret = kvm_read_guest(kvm, (gpa_t)ptr, &val, 1);
> +		if (ret)
> +			return ret;
> +		irq->pending_latch = val & (1 << (irq->intid % BITS_PER_BYTE));
> +	}
> +	return 0;

Again, this feels very similar to what its_sync_lpi_pending_table is
doing. We should be able to have some common code here.

>  }
>  
>  static int vgic_its_flush_ite(struct vgic_its *its, struct its_device *dev,
> 

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...