On Sat, May 27, 2023 at 12:18:04PM -0700, Linus Torvalds wrote: > And before you say "unlock order doesn't matter" - that's actually not > true. Unlock order does matter when you have things like > "spin_lock_irq()" where the irq-off region then protects other locks > too. So I had already verified that GCC keeps them ordered, I see the subthread with Ian and the documentation update request -- so all good there. Also, lockdep would scream bloody murder if you get that wrong, so I can certainly add something to the selftests that would detect this.
