rdma_restrict_node_type(9)

πŸ“¦ version 7.3.1, in nixpkgs unstable

πŸ’… rdma_restrict_node_type β€” bestie's guide to the Kernel API Manual ✨

okay so like, the manual name is rdma_restrict_node_type(9) and honestly? it's giving "small function, big locks-in energy" πŸŽ€


✨ NAME ✨

rdma_restrict_node_type β€” no cap, this one restricts an RDMA identifier to a specific RDMA device node type πŸ¦‹


🌸 SYNOPSIS 🌸

int rdma_restrict_node_type (struct rdma_cm_id *id , u8 node_type );

literally the whole signature, copy it exactly bestie 🫢


πŸ’– ARGUMENTS πŸ’–

  • id β€” the RDMA identifier. that's the main character right there πŸ‘‘
  • node_type β€” the device node type. and listen, only RDMA_NODE_UNSPECIFIED (that's the default, iykyk), RDMA_NODE_RNIC and RDMA_NODE_IB_CA are allowed. that's it. those three. don't try anything else πŸ’…

πŸ”₯ DESCRIPTION πŸ”₯

so what this does is it lets the caller restrict the possible devices used to iWarp (RDMA_NODE_RNIC) or InfiniBand/RoCEv1/RoCEv2 (RDMA_NODE_IB_CA). slay, very demure, very mindful 🌈

but here's the catch, and it's a big one:

it needs to be called before the RDMA identifier is bound to a device. which means it should be called before rdma_bind_addr, rdma_resolve_addr and rdma_listen. so like, before. not after. before. 😭

⭐ SEE ALSO ⭐

kernel file ./include/rdma/rdma_cm.h, and the whole friend group honestly:


okay that's the whole page bestie, go forth and restrict responsibly πŸ§šπŸ’œ

source/version: ./include/rdma  Β·  date: October 2026 β€” mwah, that's the footer, love u, bye! πŸ’…βœ¨πŸŽ€

πŸ“¦ where this page comes from