netfs_read_remote_i_size(9)

💖 NAME 💖

netfs_read_remote_i_size — bestie, this one reads remote_i_size safely, no cap 🫶

---

✨ SYNOPSIS ✨

unsigned long long netfs_read_remote_i_size (const struct inode *inode);

---

🎀 ARGUMENTS 🎀

• inode — the inode you wanna access, that's literally it 💅

---

🦋 DESCRIPTION 🦋

Okay so picture this: you just want the remote_i_size, right? And this function hands it to you safely, without any of that tearing nonsense that can happen on 32-bit arches 😭✨ Like it's giving "trust me, I've got you" energy. Iconic behavior, honestly.

---

🌸 NOTE 🌸

Alright bestie, sit down for this one because it's a whole saga 👑

So on a 32-bit arch with a preemptable kernel and an UP compile, i_size_read/i_size_write must be atomic with respect to the local CPU — and that's unlike the preempt-disabled situation. But! They don't need to be atomic with respect to other CPUs like in true SMP. So they need either to locally disable IRQ around the read, or — for example on x86 — they can still be implemented as a cmpxchg8b without needing the lock prefix 🔥

And for SMP compiles and 64-bit archs? It makes no difference whether preempt is enabled or not. Periodt. 💀

---

🌈 SEE ALSO 🌈

Kernel file ./include/linux/netfs.h — plus the whole friend group:

• folio_start_private_2(9)
• netfs_inode(9)
• netfs_read_zero_point(9)
• netfs_read_sizes(9)
• netfs_inode_init(9)
• netfs_resize_file(9)
• netfs_i_cookie(9)
• netfs_wait_for_outstanding_io(9)

---

🧚 okay that's the whole page, go touch some grass (after you read the source, obviously)

Kisses from ./include/linux, dated October 2026, and this lil page's name is netfs_read_remote_i_size(9) — catch you in the next manpage, bestie! 💜