0

I want to use RDMA in a KMDF driver where the buffer received in an EvtIoWrite will be written directly to remote memory but this operation is performance critical so I'm wondering if there is a way to avoid copying the buffer to an RDMA memory region memory first or some faster workaround I can do in kernel code that would speed it up. I have seen something called "Fast Registration Work Request" mentioned but I feel that may be a Linux only thing. An inline send would be ideal but the size of an inline send is limited to something like 928 bytes so not practical for this application.

So far nobody has even commented on my RDMA questions so I'm probly on my own with this but if you know somewhere I can get better help with RDMA please say so.

CommanderLake
  • 51
  • 9
  • Are you using NDK? Perhaps you can use [NDK_FN_FAST_REGISTER](https://learn.microsoft.com/en-us/windows-hardware/drivers/ddi/ndkpi/nc-ndkpi-ndk_fn_fast_register). – haggai_e Sep 09 '21 at 04:47
  • I'm new to kernel stuff, if I google NDK all I get is android related stuff, perhaps you could fill me in? – CommanderLake Sep 09 '21 at 13:31
  • This may loose my already non-existent audience but I use Server 2008 R2/7 because... well, I prefer it. – CommanderLake Sep 09 '21 at 13:44

0 Answers0