Re: [Qemu-devel] [PATCH v6 4/8] kvm-all: Pass requester ID to MSI routing functions

[Peter Maydell]

On 9 September 2016 at 18:30, Eric Auger <address@hidden> wrote:
> From: Pavel Fedin <address@hidden>
>
> Introduce global kvm_msi_use_devid flag and pass the device ID,
> if needed, while building the MSI route entry. Device IDs are
> required by the ARM GICv3 ITS (IRQ remapping function is based on
> this information).
>
> Signed-off-by: Pavel Fedin <address@hidden>
> Signed-off-by: Eric Auger <address@hidden>
>
> ---
> v5 -> v6:
> - move the flag in include/sysemu/kvm.h and populate the devid
>   information in kvm-all.c
>
> v3 -> v4:
> - OR route->flags with KVM_MSI_VALID_DEVID
> ---
>  include/sysemu/kvm.h | 1 +
>  kvm-all.c            | 9 +++++++++
>  kvm-stub.c           | 1 +
>  3 files changed, 11 insertions(+)
>
> diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h
> index c9c2436..c31c810 100644
> --- a/include/sysemu/kvm.h
> +++ b/include/sysemu/kvm.h
> @@ -53,6 +53,7 @@ extern bool kvm_gsi_direct_mapping;
>  extern bool kvm_readonly_mem_allowed;
>  extern bool kvm_direct_msi_allowed;
>  extern bool kvm_ioeventfd_any_length_allowed;
> +extern bool kvm_msi_use_devid;
>
>  #if defined CONFIG_KVM || !defined NEED_CPU_H
>  #define kvm_enabled()           (kvm_allowed)

You also need the appropriate #define, like all the other flags
here (with a doc comment, and with the version that's just defined
as '(false)').

The idea is that tests of the condition then don't need to
be guarded by ifdef CONFIG_KVM.

> diff --git a/kvm-all.c b/kvm-all.c
> index ebf35b0..df3bcde 100644
> --- a/kvm-all.c
> +++ b/kvm-all.c
> @@ -119,6 +119,7 @@ bool kvm_readonly_mem_allowed;
>  bool kvm_vm_attributes_allowed;
>  bool kvm_direct_msi_allowed;
>  bool kvm_ioeventfd_any_length_allowed;
> +bool kvm_msi_use_devid;
>
>  static const KVMCapabilityInfo kvm_required_capabilites[] = {
>      KVM_CAP_INFO(USER_MEMORY),
> @@ -1275,6 +1276,10 @@ int kvm_irqchip_add_msi_route(KVMState *s, int vector, 
> PCIDevice *dev)
>      kroute.u.msi.address_lo = (uint32_t)msg.address;
>      kroute.u.msi.address_hi = msg.address >> 32;
>      kroute.u.msi.data = le32_to_cpu(msg.data);
> +    if (kvm_msi_use_devid) {

these should use the macro, not direct access to the flag variable.

> +        kroute.flags = KVM_MSI_VALID_DEVID;
> +        kroute.u.msi.devid = pci_requester_id(dev);
> +    }
>      if (kvm_arch_fixup_msi_route(&kroute, msg.address, msg.data, dev)) {
>          kvm_irqchip_release_virq(s, virq);
>          return -EINVAL;
> @@ -1308,6 +1313,10 @@ int kvm_irqchip_update_msi_route(KVMState *s, int 
> virq, MSIMessage msg,
>      kroute.u.msi.address_lo = (uint32_t)msg.address;
>      kroute.u.msi.address_hi = msg.address >> 32;
>      kroute.u.msi.data = le32_to_cpu(msg.data);
> +    if (kvm_msi_use_devid) {
> +        kroute.flags = KVM_MSI_VALID_DEVID;
> +        kroute.u.msi.devid = pci_requester_id(dev);
> +    }
>      if (kvm_arch_fixup_msi_route(&kroute, msg.address, msg.data, dev)) {
>          return -EINVAL;
>      }
> diff --git a/kvm-stub.c b/kvm-stub.c
> index 64e23f6..6f896e9 100644
> --- a/kvm-stub.c
> +++ b/kvm-stub.c
> @@ -31,6 +31,7 @@ bool kvm_gsi_direct_mapping;
>  bool kvm_allowed;
>  bool kvm_readonly_mem_allowed;
>  bool kvm_ioeventfd_any_length_allowed;
> +bool kvm_msi_use_devid;
>
>  int kvm_destroy_vcpu(CPUState *cpu)
>  {

thanks
-- PMM