xiaoxiang781216 commented on PR #6834: URL: https://github.com/apache/incubator-nuttx/pull/6834#issuecomment-1212265551
> A Zero-copy ethernet solution would be indeed a great addition, when implementing a enet driver myself I noticed as well that packets are double stored and have a unnecessary copy action, The header and alignment is part of the offloading solution. The current design keep the old behavior as before, but give the implementation the possibility. > furthermore hw TCP/IP offloading (HW checksum etc) is also thing NuttX doesn't support that well. > HW checksum is already supported, you can simply enable NET_ARCH_CHKSUM: https://github.com/apache/incubator-nuttx/blob/master/net/utils/Kconfig#L6-L26 > But wouldn't it be better to come up with a proposal to offload DMA directly into a IOB buffer and start a discussion (i.e. NuttX mailing list) It's better to show the code and > how we can make this portable in such a way that it works all kinds of ethernet MAC's that have build-in DMA. The offload change is simple from the concept: 1. TCP/IP stack pass IOB directly to netdev instead copy to d_buf 2. netdev request the hardware send the data in IOB and free IOB to pool after done. 3. The netdev allocate IOB and pass to hardware for receiving 4. Pass IOB to ipv4_input/ipv6_input. TCP/IP stack will release the buffer after the data copy to the user buffer Of course, to keep compatibility, the new behavior need enabled by some flag. > I think this requires more changes into how IOB works but also the API of the NuttX Networking to work with these changes. Could you point out the detail change required for IOB? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@nuttx.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
