Re: [PATCH 0/4] Introduce vfio_dma_mapping_perf_test

[David Matlack <dmatlack@google.com>]

On 2026-07-01 08:33 PM, Aaron Lewis wrote:
> Add a new VFIO selftest, vfio_dma_mapping_perf_test, to provide a
> configurable tool for measuring DMA mapping and unmapping latencies.
> 
> This selftest introduces command-line parameters to provide a convenient
> way to tune the test for specific situations. For example, it can be used
> to observe the lengthy unmap times of the Type 1 IOMMU on large memory
> regions, or to compare DMA mapping performance between the Type 1 IOMMU
> and IOMMUFD. These scenarios can now be easily configured and run via the
> command line.
> 
> During development, I abandoned an approach where I initialized memfd
> globally in main() for all tests. While that allowed memfd to be
> initialized once for all test variants, it proved awkward because a
> different memfd would be needed for each supported page size (e.g., 4K,
> 2M, 1G). Also, If a user is only running a single test variant,
> pre-allocating all sizes is unnecessary overhead. Therefore, I dropped
> that approach in favor of test-local allocation. However, I'm open to
> ideas on a better way to handle this.
> 
> Aaron Lewis (4):
>   vfio: selftests: Introduce vfio_dma_mapping_perf_test
>   vfio: selftests: Add memfd test to vfio_dma_mapping_perf_test
>   vfio: selftests: Allow a size for vfio_dma_mapping_perf_test
>   vfio: selftests: Allow the flag MAP_POPULATE to be set on the cmdline

Please respond inline to the Sashiko feedback and address any findings
in v2. I'll be doing a review as well shortly.