Re: [PATCH v0 3/5] perf: Introduce instruction trace filtering

[Peter Zijlstra <peterz@infradead.org>]

On Fri, Dec 11, 2015 at 07:13:10PM +0200, Alexander Shishkin wrote:
> Peter Zijlstra <peterz@infradead.org> writes:
> > On second thought; we should not inherit the filters at all.
> >
> > We should always use event->parent (if exists) for filters. Otherwise
> > inherited events will get different filters if you change the filter
> > after clone.
> 
> But children will have different mappings,

_can_ have.

> so the actual filter
> configurations will still differ between parents and children. I guess I
> could split the filter in two parts: one that's defined by the user and
> one that we calculated from vma addresses, that we later program into
> hardware.

/me confused, isn't that what you already do?

In any case, since inherited counters are uncontrollable (they have no
filedesc of their own) and you cannot a priory tell what a child will go
do, let alone a child of a child. It really makes no sense to have
different filters on different parts of the inherited tree.