Re: [PATCH 3/3] KVM: selftests: Use VFIO selftests library to pick IOMMU mode in IRQ test

[Sean Christopherson <seanjc@google.com>]

On Thu, Oct 29, 2026, David Matlack wrote:
> Use the VFIO selftests library to pick the IOMMU mode in the IRQ test and
> drop probe_iommu_type(). Pass NULL to iommu_init() when the user does not
> specify a mode with -t, and skip the test if the requested mode (or, if
> NULL, any mode) is not available.
> 
> The VFIO selftests library now automatically picks the IOMMU mode that is
> most likely to work on the current system, so there is no need for each
> test to probe for a mode itself.
> 
> Suggested-by: Sean Christopherson <seanjc@google.com>
> Link: https://lore.kernel.org/kvm/aiIG2L_KdPwuxYA1@google.com/
> Signed-off-by: David Matlack <dmatlack@google.com>
> ---
>  tools/testing/selftests/kvm/irq_test.c | 19 +------------------
>  1 file changed, 1 insertion(+), 18 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/irq_test.c b/tools/testing/selftests/kvm/irq_test.c
> index 791e576c334c..f9c11d80a306 100644
> --- a/tools/testing/selftests/kvm/irq_test.c
> +++ b/tools/testing/selftests/kvm/irq_test.c
> @@ -148,22 +148,6 @@ static void kvm_set_empty_gsi_routing(struct kvm_vm *vm)
>  	vm_ioctl(vm, KVM_SET_GSI_ROUTING, &routing);
>  }
>  
> -static const char *probe_iommu_type(void)
> -{
> -	int io_fd;
> -
> -	io_fd = open("/dev/iommu", O_RDONLY);
> -	if (io_fd >= 0) {
> -		close(io_fd);
> -		return MODE_IOMMUFD;
> -	}
> -
> -	io_fd = __open_path_or_exit("/dev/vfio/vfio", O_RDONLY,
> -				    "Is VFIO (or IOMMUFD) loaded and enabled?");
> -	close(io_fd);
> -	return MODE_VFIO_TYPE1_IOMMU;
> -}
> -
>  static void help(const char *name)
>  {
>  	printf("Usage: %s [-a] [-d <segment:bus:device.function>] [-e] [-h] [-i nr_irqs] [-m] [-n] [-t iommu_type] [-v nr_vcpus] [-x]\n", name);
> @@ -266,8 +250,7 @@ int main(int argc, char **argv)
>  	}
>  
>  	if (device_bdf) {
> -		if (!iommu_type)
> -			iommu_type = probe_iommu_type();
> +		TEST_REQUIRE(iommu_mode_is_available(iommu_type));

This is very unhelpful when neither IOMMUFD nor VFIO is available, especially
why the user doesn't explicitly provide the desired mode:

  # ./irq_test -d 0000:6f:01.0 -i 10000
  Random seed: 0x79d386bd
  1..0 # SKIP - Requirement not met: iommu_mode_is_available(iommu_type)

I helped write the test and I had to go read the code to figure out what iommu_type
would be, and then read the VFIO library to see how it reacted to a NULL pointer.

How about something like this as fixup:

diff --git a/tools/testing/selftests/kvm/irq_test.c b/tools/testing/selftests/kvm/irq_test.c
index f9c11d80a306..63205afc04d1 100644
--- a/tools/testing/selftests/kvm/irq_test.c
+++ b/tools/testing/selftests/kvm/irq_test.c
@@ -250,7 +250,9 @@ int main(int argc, char **argv)
 	}
 
 	if (device_bdf) {
-		TEST_REQUIRE(iommu_mode_is_available(iommu_type));
+		__TEST_REQUIRE(iommu_mode_is_available(iommu_type),
+			       "IOMMU mode(s) '%s' not available",
+			       iommu_type ? iommu_type : "IOMMU and/or VFIO");
 		iommu = iommu_init(iommu_type);
 		device = vfio_pci_device_init(device_bdf, iommu);
 		msix = vfio_setup_msi(device);

to yield:

  # ./irq_test -d 0000:6f:01.0 -i 10000 -t vfio_type1_iommu
  Random seed: 0x14e21328
  1..0 # SKIP - IOMMU mode(s) 'vfio_type1_iommu' not available
  # ./irq_test -d 0000:6f:01.0 -i 10000
  Random seed: 0x160225ae
  1..0 # SKIP - IOMMU mode(s) 'IOMMU and/or VFIO' not available