spin_lock_bh() --> BH disable in hard IRQ, OH NOEZ!

No I didn't see his original message just the patch - his original is a
fixed bug.

mkiss_receive_buf(), the problematic code path, is invoked by
the TTY layer directly from LDISC flushing and the TTY device's
IRQ handler here.

How does networking's TX policy have anything to do with this?

Sorry I'd been looking at the other side stuff with ppp which is a
sleeping v bh issue and misinterpreted Ralf's post.

In fact Ralf's fix isn't needed. The incorrect uses of tty->low_latency
all got killed and the pty code got rewritten which was the other case
that could occur.