hfi1_init(9)

💅 hfi1_init(9) — like, the actual chip init era

Okay bestie, so this is a page from the Kernel API Manual, and the star of the show is hfi1_init(9). Buckle up, we're going in 🎀

---

✨ NAME ✨

hfi1_init — do the actual initialization sequence on the chip 🧚

---

🌸 SYNOPSIS 🌸

int hfi1_init (struct hfi1_devdata *dd , int reinit );

---

🦋 ARGUMENTS 🦋

• dd — the hfi1_ib device. the whole entire vibe of a device. the main character 💅
• reinit — re-initializing, so don't allocate new memory. like bestie, we already have memory at home 🙈

---

🔥 DESCRIPTION 🔥

Do the actual initialization sequence on the chip. iconic behaviour, no notes 👑

And it's not a one-time thing either — this happens from the init routine called from the PCI infrastructure, and whenever we reset the chip, or when we detect that it was reset internally, or when it's administratively re-enabled. so many entrances, honestly 🌈

Memory allocation here and in called routines is only done in the first case, i.e. when reinit == 0. periodt.

okay so this is the careful part, pay attention bestie 😭 — we have to be careful, because even without memory allocation, we still need to re-write all the chip registers, TIDs, etc. after the reset or enable has completed. like, all of it. the whole registry glow-up. don't skip it or the vibe check fails fr fr 💀

---

💜 SEE ALSO 💜

Kernel file ./drivers/infiniband/hw/hfi1/init.c, plus the whole friend group of related calls — go say hi 🫶

• hfi1_rcd_free(9)
• hfi1_rcd_put(9)
• hfi1_rcd_get(9)
• allocate_rcd_index(9)
• hfi1_rcd_get_by_index_safe(9)
• hfi1_rcd_get_by_index(9)
• hfi1_free_ctxt(9)
• init_after_reset(9)
• create_workqueues(9)
• destroy_workqueues(9)
• enable_general_intr(9)
• shutdown_device(9)
• hfi1_free_ctxtdata(9)
• hfi1_free_devdata(9)
• hfi1_alloc_devdata(9)
• hfi1_create_rcvhdrq(9)
• hfi1_setup_eagerbufs(9)

---

✨ mwah ✨ that's the whole page, bestie — brought to you straight outta ./drivers/infiniband/hw/hfi1, dated October 2026. go forth and init responsibly 💖🎀🦋