md_write_metadata(9)

💾 md_write_metadata(9) — bestie edition

okay so this is a page from the Kernel API Manual ✨ and it's literally about md_write_metadata — the little function that gets your metadata onto the actual disk. buckle up 🎀

---

🌸 NAME 🌸

md_write_metadata — write metadata to underlying disk, including array superblock, badblocks, bitmap superblock and bitmap bits. that's the whole menu, bestie 💅

---

✨ SYNOPSIS ✨

void md_write_metadata (struct mddev *mddev , struct md_rdev *rdev ,
sector_t sector , int size , struct page *page , unsigned int offset );

no cap, that's the whole signature. six args. she's thorough 👑

---

🦋 ARGUMENTS 🦋

• mddev — the array to write 💖 this is the big boss array itself.
• rdev — the underlying disk to write 🫶 the actual physical disk doing the heavy lifting.
• sector — the offset to rdev ⭐ where on that disk we're landing.
• size — the length of the metadata 🌸 how many bytes of the stuff we're pushing out.
• page — the metadata 🎀 the actual payload, sitting pretty in a page.
• offset — the offset to page 🧚 where inside that page our data starts.

---

🔥 DESCRIPTION 🔥

okay so here's the vibe: it writes size bytes of page starting from offset, out to sector of rdev. that's the whole core move, iykyk 💅

then it increments mddev->pending_writes before returning, and decrements it on completion, waking up sb_wait 😭 yes she's tracking herself, it's called accountability.

⚠️ important, bestie: the caller must call md_super_wait after issuing io to all rdev. not optional. like, do it. 🙈

and if an error occurred, md_error will be called, and the rdev will be kicked out from mddev 💀 yeah. eviction. brutal but fair.

---

🌈 SEE ALSO 🌈

• Kernel file ./drivers/md/md.c
• stop_sync_thread(9)

---

that's the whole page, bestie — go write some metadata, slay 💜

source: ./drivers/md · October 2026 · md_write_metadata(9) — mwah, see u in the group chat 🫶✨