On Wed, Aug 3, 2022 at 2:55 PM Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> The original patch years ago use to have:
>
> preempt_disable_rt()
>
> preempt_enable_rt()
That may be visually simpler, but I dislike how it's named for some
implementation detail, rather than for the semantic meaning.
Admittedly I think "preempt_enable_under_spinlock()" may be a bit
*too* cumbersome as a name. It does explain what is going on - and
both the implementation and the use end up being fairly clear (and the
non-RT case could have some debug version that actually tests that
preemption has already been disabled).
But it is also a ridiculously long name, no question about that.
I still feel is less cumbersome than having that
"IS_ENABLED(CONFIG_PREEMPT_RT)" test that also then pretty much
requires a comment to explain what is going on.
Linus
