On Fri, 11 Oct 2019 09:37:10 +0200
Daniel Bristot de Oliveira <bris...@redhat.com> wrote:

> But, yes, we will need [ as an optimization ] to sort the address right before
> inserting them in the batch. Still, having the ftrace_pages ordered seems to 
> be
> a good thing, as in many cases, the ftrace_pages are disjoint sets.

I think it would be fine if we run the batches according to the ftrace
page groups. Which will be guaranteed to be sorted. I try to keep the
groups to a minimum, thus it shouldn't be too many ipi busts.

Although, my new work may actually make more page groups by trying to
cut down on the dyn_ftrace size by using offsets. If a function extends
outside the offset range, a new page group will need to be created.

-- Steve

Reply via email to