blk_rq_dma_unmap(9)

💅 blk_rq_dma_unmap — the Kernel API Manual tea 🎀

okay so bestie, welcome to the blk_rq_dma_unmap(9) page from the Kernel API Manual — yes it's a real kernel man page, yes we're gonna make it cute, and no we are NOT skipping a single detail. periodt. ✨

---

🌸 NAME 🌸

blk_rq_dma_unmap — try to DMA unmap a request

like, that's the whole vibe. it tries. it's a *try*er. 🫶

---

⭐ SYNOPSIS ⭐

bool blk_rq_dma_unmap (struct request *req , struct device *dma_dev ,
struct dma_iova_state *state , size_t mapped_len , enum
pci_p2pdma_map_type map );

and it returns a bool, bestie. that bool is the main character here, iykyk. 👑

---

🦋 ARGUMENTS 🦋

literally every single one, no skips:

• req — the request to unmap 💌
• dma_dev — the device to unmap from 🌈
• state — DMA IOVA state, iconic little state machine energy ✨
• mapped_len — number of bytes to unmap, yes in bytes, ngl be precise 🍓
• map — peer-to-peer mapping type 🧚

---

💜 DESCRIPTION 💜

okay so here's the tea, and it's genuinely important so listen up:

it returns false if the callers need to manually unmap every DMA segment mapped using iter — like, you gotta do the work yourself, sorry bestie 😭 — or true if no work is left to be done. that's it. that's the whole function. clean. 💅

like the way that a single bool can carry this much meaning?? obsessed. 🔥

---

🌈 SEE ALSO 🌈

go touch grass in these, respectively:

• ./include/linux/blk-mq-dma.h — a kernel file, go read it 💖
• blk_rq_dma_map_coalesce(9) — the sibling page, the lore continues 🎀

---

fr fr: don't sleep on the false branch. if it says false, you are the one unmapping every segment. no cap. 🙈

---

okay bye bestie, stay slay and stay mapped (until you're not) 💖✨🦋

— from ./include/linux, dated October 2026, blk_rq_dma_unmap(9), signed off with a little kiss 🎀