[PATCH v9 2/9] preempt: Introduce __preempt_count_{sub, add}_return()

Peter Zijlstra peterz at infradead.org
Fri Feb 28 01:24:05 PST 2025


On Fri, Feb 28, 2025 at 10:15:09AM +0100, Heiko Carstens wrote:

> I couldn't find any cover letter for the whole patch series which describes
> what this is about, and why it is needed.
> It looks like some Rust enablement?

Yeah, more or less.

It's replacing local_irq_save() and all related functions
(spin_lock_irqsave etc..) that take a flags argument with this new thing
that frobs a recursion count in preempt_count(), obviating the need to
carry the local flags argument around.

This is nice, even for C code, less flags muck to carry around.

It would be even better if they then went and deleted all of the _irq /
_irqsave nonsense entirely.

Yes, that's going to be a big patch :-)

Also, IIRC there is some arch stuff that comes unstuck if you do this
blindly (I tried at some point, it didn't boot).



More information about the linux-arm-kernel mailing list