On 23/03/15 15:58, Andre Przywara wrote:
> Using the framework provided by the recent vgic.c changes, we
> register a kvm_io_bus device on mapping the virtual GICv3 resources.
> The distributor mapping is pretty straight forward, but the
> redistributors need some more love, since they need to be tagged with
> the respective redistributor (read: VCPU) they are connected with.
> We use the kvm_io_bus framework to register two devices per VCPU, as
> each block is handled independently by the VGIC code.
>
> Signed-off-by: Andre Przywara <andre.przyw...@arm.com>
> ---
> include/kvm/arm_vgic.h | 1 +
> virt/kvm/arm/vgic-v3-emul.c | 34 +++++++++++++++++++++++++++++++++-
> virt/kvm/arm/vgic.c | 18 ++++++++++++++++++
> 3 files changed, 52 insertions(+), 1 deletion(-)
>
> diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h
> index 4523984..d6705f4 100644
> --- a/include/kvm/arm_vgic.h
> +++ b/include/kvm/arm_vgic.h
> @@ -252,6 +252,7 @@ struct vgic_dist {
>
> struct vgic_vm_ops vm_ops;
> struct vgic_io_device dist_iodev;
> + struct vgic_io_device *redist_iodevs;
> };
>
> struct vgic_v2_cpu_if {
> diff --git a/virt/kvm/arm/vgic-v3-emul.c b/virt/kvm/arm/vgic-v3-emul.c
> index 14943e3..35679d1 100644
> --- a/virt/kvm/arm/vgic-v3-emul.c
> +++ b/virt/kvm/arm/vgic-v3-emul.c
> @@ -766,6 +766,9 @@ static int vgic_v3_map_resources(struct kvm *kvm,
> {
> int ret = 0;
> struct vgic_dist *dist = &kvm->arch.vgic;
> + gpa_t rdbase = dist->vgic_redist_base;
> + struct vgic_io_device *iodevs = NULL;
> + int i;
>
> if (!irqchip_in_kernel(kvm))
> return 0;
> @@ -791,7 +794,36 @@ static int vgic_v3_map_resources(struct kvm *kvm,
> goto out;
> }
>
> - kvm->arch.vgic.ready = true;
> + ret = vgic_register_kvm_io_dev(kvm, dist->vgic_dist_base,
> + GIC_V3_DIST_SIZE, vgic_v3_dist_ranges,
> + -1, &dist->dist_iodev);
> + if (ret)
> + goto out;
> +
> + iodevs = kcalloc(dist->nr_cpus * 2, sizeof(iodevs[0]), GFP_KERNEL);
> + if (!iodevs) {
> + ret = -ENOMEM;
> + goto out;
> + }
> +
> + /* kvm_vgic_destroy() will take care of destroying the devices later. */
> + for (i = 0; i < dist->nr_cpus; i++) {
> + ret = vgic_register_kvm_io_dev(kvm, rdbase,
> + SZ_64K, vgic_redist_ranges,
> + i, &iodevs[i * 2]);
> + if (ret)
> + goto out;
> + ret = vgic_register_kvm_io_dev(kvm, rdbase + SGI_BASE_OFFSET,
> + SZ_64K, vgic_redist_sgi_ranges,
> + i, &iodevs[i * 2 + 1]);
> + if (ret)
> + goto out;
> + rdbase += GIC_V3_REDIST_SIZE;
> + }
The "two devices per redistributor" sent me spinning for a while. I can
see why you're doing so (it makes it easier to switch from one model to
the other), but I'd really like to see a subsequent patch that merge them.
It will be a lot more readable, and will match the architecture (despite
being on two separate 64k pages, they are part of the same device).
> +
> + dist->redist_iodevs = iodevs;
> + dist->ready = true;
> +
> out:
> if (ret)
> kvm_vgic_destroy(kvm);
> diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
> index 9a732d0..9cbb55f4 100644
> --- a/virt/kvm/arm/vgic.c
> +++ b/virt/kvm/arm/vgic.c
> @@ -992,6 +992,8 @@ int vgic_register_kvm_io_dev(struct kvm *kvm, gpa_t base,
> int len,
> static void vgic_unregister_kvm_io_dev(struct kvm *kvm)
> {
> struct vgic_dist *dist = &kvm->arch.vgic;
> + struct vgic_io_device *iodevs;
> + int i;
>
> if (!dist || !kvm->buses[KVM_MMIO_BUS])
> return;
> @@ -1001,6 +1003,22 @@ static void vgic_unregister_kvm_io_dev(struct kvm *kvm)
> if (dist->dist_iodev.dev.ops)
> kvm_io_bus_unregister_dev(kvm, KVM_MMIO_BUS,
> &dist->dist_iodev.dev);
> +
> + iodevs = dist->redist_iodevs;
> + if (iodevs) {
> + for (i = 0; i < dist->nr_cpus * 2; i++) {
> + /*
> + * Because of a failed initialization we could get here
> + * without ever having registered a device.
> + */
> + if (!iodevs[i].dev.ops)
> + continue;
> + kvm_io_bus_unregister_dev(kvm, KVM_MMIO_BUS,
> + &iodevs[i].dev);
> + }
> + kfree(iodevs);
> + dist->redist_iodevs = NULL;
> + }
> mutex_unlock(&kvm->slots_lock);
> }
>
>
Other than the above,
Reviewed-by: Marc Zyngier <marc.zyng...@arm.com>
M.
--
Jazz is not dead. It just smells funny...
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
