Allow passing a dma_device when creating the virtqueue through a new Let vring_dma_device() return the per virtqueue dma device instead Some of the work are actually done before since the helper like One example is the vDPA that where the control virtqueue could be This will be usedįor virtio devices whose virtqueue are backed by different underlayer This patch introduces a per virtqueue dma device. ` (4 subsequent siblings) 5 siblings, 2 replies 22+ messages in thread 6:28 ` vdpa: introduce get_vq_dma_device() Jason Wang 6:28 virtio_ring: per virtqueue DMA device Jason 6:28 ` Jason Wang So DMA API can go with the direct mapping then the vringh willīe happy since mlx5_vdpa assuems a direct/identical mapping by This isįixed by advetise the vDPA device that doesnt do DMA (without a DMA IOMMU but the vringh assumes a direct mapping (PA as IOVA). That the virito_ring tries to map the control virtqueue into platform This fixes the crash of mlx5_vdpa + virtio_vdpa when platform IOMMU isĮnabled but not in the passthrough mode. Software virtqueues can utilize the transport specific method to So this series tries to introduce per virtqueue DMA device, then Physical DMA device for software emulated device, since there's noĮasy way for vringh to know about the hardware IOMMU mappings. In this case, it would be wrong if we stick with the OneĮxample is that in the case of vDPA some parent may emualte virtqueue In some cases, the virtqueue could be backed by different devices. To: mst, jasowang +Cc: elic, gdawar, virtualization, linux-kernel, tanuj.kamde ` (5 more replies) 0 siblings, 6 replies 22+ messages in thread 6:28 ` virtio_ring: per virtqueue dma device Jason Wang Virtio_ring: per virtqueue DMA device LKML Archive on help / color / mirror / Atom feed * virtio_ring: per virtqueue DMA device 6:28 Jason Wang
0 Comments
Leave a Reply. |