On 2024/4/19 20:56, Jason Gunthorpe wrote:
+ riscv_iommu_cmd_inval_vma(&cmd); + riscv_iommu_cmd_inval_set_pscid(&cmd, domain->pscid); + if (len > 0 && len < RISCV_IOMMU_IOTLB_INVAL_LIMIT) { + for (iova = start; iova < end; iova += PAGE_SIZE) { + riscv_iommu_cmd_inval_set_addr(&cmd, iova); + riscv_iommu_cmd_send(iommu, &cmd, 0); + } + } else { + riscv_iommu_cmd_send(iommu, &cmd, 0); + } + }This seems suboptimal, you probably want to copy the new design that Intel is doing where you allocate "bonds" that are already de-duplicated. Ie if I have 10 devices on the same iommu sharing the domain the above will invalidate the PSCID 10 times. It should only be done once. ie add a "bond" for the (iommu,pscid) and refcount that based on how many devices are used. Then another "bond" for the ATS stuff eventually.
The latest version is under discussion here. https://lore.kernel.org/linux-iommu/20240416080656.60968-1-baolu.lu@xxxxxxxxxxxxxxx/ Supposedly, you can make such optimization after the base code is landed in the mainline tree if the change is big. Best regards, baolu
