On Wed, Aug 3, 2022 at 4:24 PM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>
> On Wed, Aug 03, 2022 at 06:59:36PM -0400, Steven Rostedt wrote:
> >
> > preempt_disable_inlock() ?
>
> preempt_disable_locked()?
Heh. Shed painting in full glory.
Let's try just "preempt_enable_under_spinlock()" and see.
It's a bit long, but it's still shorter than the existing usage pattern.
And we don't have "inlock" anywhere else, and while "locked" is a real
pattern we have, it tends to be about other things (ie "I hold the
lock that you need, so don't take it").
And this is _explicitly_ only about spinning locks, because sleeping
locks don't do the preemption disable even without RT.
So let's make it verbose and clear and unambiguous. It's not like I
expect to see a _lot_ of those. Knock wood.
We had a handful of those things before (in mm/vmstat, and now added
another case to the dentry code. So it has become a pattern, but I
really really hope it's not exactly a common pattern.
And so because it's not common, typing a bit more is a good idea - and
making it really clear is probably also a good idea.
Linus
