iscsi_iser_conn_stop(9)

📦 version 7.3.1, in nixpkgs unstable

💅 iscsi_iser_conn_stop(9) — the Kernel API Manual moment

okay so bestie, welcome to the glow-up era of the kernel API manual for iscsi_iser_conn_stop(9) 🎀 this one's short but she matters, so like, pay attention fr fr.

✨ NAME ✨

iscsi_iser_conn_stop — stop an iscsi-iser connection 💅 like, the connection gets told to wrap it up, that's the whole vibe.

🌸 SYNOPSIS 🌸

void iscsi_iser_conn_stop (struct iscsi_cls_conn *cls_conn , int flag );

that's it. that's the whole signature. no cap. 🫶

🦋 ARGUMENTS 🦋

  • cls_conn — an iscsi class connection 👑 like, the connection object itself, the main character of this function.
  • flag — tells you whether this is a recover or a terminate 💀 and here's the thing: it gets passed as is. no filtering, no judging, no little cleanup pass. it goes straight through. periodt.

🔥 NOTES 🔥

okay this is the part where I need you to lock in 🙈 because there's a real caveat here.

Calling iscsi_conn_stop might theoretically race with a DEVICE_REMOVAL event and dereference a previously freed RDMA device handle 😭 and that is exactly the kind of chaos we're not doing.

So: we call it under the iser state lock, and that's specifically to protect against this kind of race 💜 the lock is the whole point, she's the bodyguard, don't skip her.

ngl the word "theoretically" is doing a lot of work here — but the fix is real and you should respect it. 🧚

🌈 SEE ALSO 🌈

kernel file ./drivers/infiniband/ulp/iser/iscsi_iser.c — go read it, it's the source of truth ⭐

and the whole friend group 💖:

and that's the page bestie 🎀 from ./drivers/infiniband/ulp/iser, dated October 2026 — slay, go stop some connections responsibly ✨💅

📦 where this page comes from