irdma_sc_ge...d_size_gen_3(9)

📦 version 7.3.1, in nixpkgs unstable

💅 irdma_sc_ge...d_size_gen_3(9) — straight out of the Kernel API Manual 🎀

okay bestie, welcome to the Kernel API Manual entry for irdma_sc_ge...d_size_gen_3(9) — yes the title is literally abbreviated with dots, that's just how the header line renders it, we move. ✨

🌸 NAME 🌸

irdma_sc_get_encoded_ird_size_gen_3 — get encoded IRD size for GEN 3 🦋

no cap, that's the whole job. it takes an IRD size and gives you back the encoded one for GEN 3.

✨ SYNOPSIS ✨

u8 irdma_sc_get_encoded_ird_size_gen_3 (u16 ird_size );

so it hands you back a u8, and you feed it a u16 named ird_size. tiny function, big main-character energy 👑

💖 ARGUMENTS 💖

  • ird_size — the IRD size, bestie 💜 so here's the tea: the ird that comes in from the connection gets rounded to a supported HW setting, and then it gets encoded for the ird_size field of qp_ctx. that's the whole pipeline. and like — consumers are expected to provide a valid ird size based on hardware attributes, so pls do your homework 🙈. if the input is invalid? IRD size defaults to a value of 4. that's the fallback, that's the safety net, that's the vibe. 🫶
ngl, "consumers are expected to provide valid ird size based on hardware attributes" is the manpage equivalent of "you should already know this" 😭 read it, respect it, pass valid input.

🌈 SEE ALSO 🌈

the kernel file lives at ./drivers/infiniband/hw/irdma/ctrl.c — that's the one, that's the home base 🏠

and here's the whole extended friend group, every single one, no skips 💅:

like, that is an iconic amount of cross-references, and yes you should go read them all 🧚 iykyk

okay that's the whole page bestie, go forth and encode your IRD sizes responsibly 💜 — this cutie came out of ./drivers/infiniband/hw/irdma and is dated October 2026, straight from the Kernel API Manual. periodt. 🎀✨

📦 where this page comes from