On Fri, Jun 13, 2025 at 4:28 AM Kyo.Liu <kyo....@nebula-matrix.com> wrote: > > Hello, Stephen: > Thanks for your review. This patch is particularly important for > our NBL driver because our coexistence implementation depends on it.
"important" is different from "required". It seems safer to plan for an integration in DPDK without this dependency, then add it later. > Unlike RDMA devices that support hardware page table translation, our > hardware requires a software-managed approach to avoid IOVA conflicts since > both user-space and kernel-space DMA operations share the same domain. This > architectural constraint makes the patch essential for: > 1. Maintaining correct memory isolation > 2. Preventing IOVA collisions in shared DMA operations > 3. Enabling safe driver coexistence > > We'd be happy to: > 1. Provide additional technical details if needed > 2. Make any necessary adjustments based on your feedback Is the kernel driver being upstreamed? -- David Marchand
