__sdma_txclean(9)

✨ __sdma_txclean ✨

okay so bestie, we're in the kernel API manual today, section 9 of the fine manual pages, and this one is called __sdma_txclean — it cleans tx of mappings, descp *kmalloc's. yes that asterisk is in the original name line, we don't ask questions, we just slay 💅

---

🌸 SYNOPSIS 🌸

void __sdma_txclean (struct hfi1_devdata *dd , struct sdma_txreq *tx );

literally a void function with two args, no return value, she just does her thing and leaves ✨

---

🎀 ARGUMENTS 🎀

• dd — the hfi1_devdata for unmapping. this is the handle she needs so she knows what to unmap 👑
• tx — the tx request to clean. this is the little guy getting tidied up 🧚

---

💖 DESCRIPTION 💖

so the way this works: this is used in the progress routine to clean the tx, or by the ULP to toss an in-process tx build. both of those vibes, one function, iconic 🫶

and the bestie-est part? the code can be called multiple times without issue. no cap. call her once, call her five times, she does not care, she is unbothered 😌

fr fr though, that re-entrancy-ish "call it as many times as you want" note is the whole personality of this one — don't skip it 🙈

---

🦋 SEE ALSO 🦋

she lives in the kernel file ./drivers/infiniband/hw/hfi1/sdma.c, and here's the whole friend group:

• sdma_state_name(9)
• sdma_get_descq_cnt(9)
• sdma_engine_get_vl(9)
• sdma_select_engine_vl(9)
• sdma_select_engine_sc(9)
• sdma_seqfile_dump_cpu_list(9)
• sdma_map_init(9)
• sdma_clean(9)
• sdma_init(9)
• sdma_all_running(9)
• sdma_start(9)
• sdma_exit(9)
• sdma_engine_error(9)
• sdma_seqfile_dump_sde(9)
• sdma_send_txreq(9)
• sdma_send_txlist(9)
• sdma_ahg_alloc(9)
• sdma_ahg_free(9)
• _sdma_engine_progress_schedule(9)

like, the entire sdma cast showed up to the group chat 💀 obsessed honestly 🌈

---

⭐ the sign-off ⭐

okay that's the whole page, mwah 💜 this one ships out of ./drivers/infiniband/hw/hfi1 and the date on the tin is October 2026 — go clean your tx, bestie, and remember: multiple calls, no issue, periodt 🎀🌸✨